CN106840161A

CN106840161A - Air navigation aid and device

Info

Publication number: CN106840161A
Application number: CN201611170559.4A
Authority: CN
Inventors: 霍光磊; 王瑾琨; 常元章; 严洁; 易梅
Original assignee: Hna Ecology Technology Group Co Ltd
Current assignee: Hna Ecology Technology Group Co Ltd
Priority date: 2016-12-16
Filing date: 2016-12-16
Publication date: 2017-06-13

Abstract

The invention discloses a kind of air navigation aid and device.Wherein, the method includes：Obtain starting point semantic region and target semantic region in predetermined space, wherein, the predetermined space is divided into multiple semantic regions, the multiple semantic region includes the starting point semantic region and the target semantic region, and the feature for identifying the semantic region from the multiple semantic region is included in each semantic region；Based on semantic region planning from the starting point semantic region to the path of the target semantic region；Equipment is set to move to the target semantic region from the starting point semantic region according to the path.The present invention solves the too wide in range technical problem of navigational semantic scope.

Description

Navigation method and device

Technical Field

The invention relates to the field of navigation, in particular to a navigation method and a navigation device.

Background

In 2005, Galindo et al constructed a two-layer map of a spatial information layer and a semantic information layer, each layer of map being associated by an "anchor". Their research is focused on semantic information, data collected by sensors (such as image data or grid information) making information of spatial information layers meaningful. Indoor real objects are sources of semantic information, and in a map designed by the indoor real objects, the semantic information is given artificially, and although the map does not have flexibility, the semantic information is easy to realize.

Vasudevan and the like design a full probability semantic map, and semantic information is realized through object identification. Objects contained in the indoor environment are converted into abstractions of space and semantics. The robot converts the space abstraction into a concept and identifies objects using semantic abstraction. And training the object set to obtain the concept. The test stage divides the space for object detection. Information of objects in an indoor environment is used to classify space, but the class information cannot be used to infer indoor objects.

Meger et al achieve automatic detection of indoor objects and form a geometric map with object information. They implement a more advanced system and implement a vision subsystem. The map construction is realized by FastSLAM, and the model of the object is distinguished by training according to image data on the Internet. The system has the advantages that indoor objects are automatically detected, and the service robot can automatically acquire semantic information in the room. Viswanathan et al propose a system with better robustness and automation, and their systems use semantic tags of entity information to extract and identify spatial semantics. The system data are also acquired from the Internet, a Bayesian model is constructed according to the frequency of the detected entities, and the experimental result of the method shows that the system can stably search typical entities and reason typical spatial semantics. The process is more stable than the above mentioned process.

Hertzberg et al propose object-based spatial semantic maps. Their objects are obtained by 3D information and a 6DSLAM method is implemented by lidar. The method first obtains point cloud information and then constructs rough features such as floors, walls, etc. On the basis, the object is detected through the classifier, and the object is projected into the map. The finally obtained information is beneficial to observation and the realization of augmented reality. The method relates to point cloud information, the processed data volume is high, and the calculation complexity is high.

Aiming at the problem that the navigation semantic scope is too wide, an effective solution is not provided at present.

Disclosure of Invention

The embodiment of the invention provides a navigation method and a navigation device, which at least solve the technical problem that the navigation semantic scope is too wide.

According to an aspect of an embodiment of the present invention, there is provided a navigation method, including: acquiring a starting point semantic region and a target semantic region in a predetermined space, wherein the predetermined space is divided into a plurality of semantic regions, the plurality of semantic regions comprise the starting point semantic region and the target semantic region, and each semantic region comprises a feature for identifying the semantic region from the plurality of semantic regions; planning a path from the starting semantic region to the target semantic region based on a semantic region; causing a device to move from the start semantic area to the target semantic area according to the path.

Further, causing the device to move from the origin semantic area to the target semantic area according to the path comprises: identifying the obtained image through the equipment to obtain a semantic area corresponding to the image, wherein the semantic area identified by the equipment is used for determining the semantic area where the equipment is located at present; and moving the equipment to the target semantic area according to the semantic area where the equipment is located at present and the path.

Further, identifying, by the device, the obtained image to obtain a semantic region corresponding to the image includes: acquiring feature points in the image; matching the feature points in the image with the feature points in a feature point library corresponding to each preset semantic region, wherein one or more feature points corresponding to each semantic region are recorded in the feature point library; determining whether the number of feature points in the image matched with the feature points in the feature point library corresponding to each semantic region meets a preset condition; and determining a semantic area corresponding to the image under the condition that the preset condition is met.

Further, acquiring feature points in the image includes: determining whether the number of feature points in the image belongs to a predetermined range; and acquiring the characteristic point in the image under the condition of belonging to a preset range.

Further, determining whether the number of feature points in the image that match feature points in the feature point library satisfies a predetermined condition comprises: comparing the matching quantity of the feature points in the image with the feature points in the feature point library corresponding to each semantic region with a preset threshold value; determining that the number of matches is greater than a predetermined threshold meeting the predetermined condition.

Further, determining whether the number of feature points in the image that match feature points in the feature point library satisfies a predetermined condition comprises: according to the matching number of the feature points in the image and the feature points in the feature point library corresponding to each semantic region, sequencing each semantic region according to the sequence of the matching number from high to low; determining that the semantic region with the top ranking order meets a preset condition.

Further, planning a path from the start semantic region to the target semantic region based on semantic regions comprises: obtaining a semantic relation in a task instruction based on semantic analysis of the task instruction; and determining the starting point semantic region and the target semantic region according to the semantic relation, and planning a path from the starting point semantic region to the target semantic region.

Further, planning a path from the start semantic region to the target semantic region based on semantic regions comprises: mapping the semantic area in the preset space into a regular graph block on a map corresponding to the preset space; and planning the path according to the rule graphic block corresponding to the semantic area.

Further, planning the path according to the rule graph block corresponding to the semantic area includes: planning a path from the starting point semantic region to the target semantic region according to a Bayesian inference model.

Further, in the case that the rule graph block is a rectangle and the device can only move in four directions, namely up, down, left and right, in the rectangle, the model of bayesian inference is as follows: o denotes the four directions of the rectangle,indicating that the device is capable of moving in an upward direction of the semantic region N in the predetermined space R,indicating that the device is capable of moving in a downward direction of the semantic region N in the predetermined space R,indicating that the device is capable of moving in the leftward direction of the semantic region N in the predetermined space R,indicating that the device is capable of moving in a rightward direction of the semantic region N in the predetermined space R.

Further, in a case that the regular image block is a rectangle and the device is able to determine, in the rectangle, the number of rotations of the device from a certain semantic region to a target semantic region in a case that each rotation direction is 90 °, wherein the number of rotations is determined according to a previous movement direction of the device to a previous semantic region and a current movement direction of the device to a current semantic region, the bayesian inference model includes:representing the number of turns of the device to move from a previous semantic area to a current semantic area, o_iRepresents the last direction of movement of the device, the o_jRepresents the current direction of movement of the device, the o_iAnd said o_jThe four moving directions of the upper, the lower, the left and the right can be respectively taken; if the last moving direction o_iAnd the current moving direction o_jSame, then the number of rotationsIf the last moving direction o_iAnd the current moving direction o_jBy 90 deg., the number of rotationsIf the last moving direction o_iAnd the current moving direction o_jBy 180 deg., the number of rotations

Further, the transition probability is a probability that the device moves to the current moving direction of the current semantic region after moving to the previous moving direction of the previous semantic region, wherein the previous semantic region and the current semantic region are adjacent semantic regions; the transition probability is calculated according to the number of turns of the device moving from the previous semantic region to the current semantic region and all rotation directions of the device, and the Bayesian inference model comprises the following steps: the formula for the label transition probability is interpreted as:wherein,representing the probability that the device moves to the current movement direction of the current semantic region again after moving to the previous movement direction of the previous semantic region, wherein,representing the transition probability, R_MAnd R_NAre two adjacent semantic regions in the predetermined space; the calculation formula of the transition probability is as follows:wherein omega_oIndicating the overall rotational direction of the device.

Further, obtaining the same preset according to the recursion relation of the Bayesian network and the transition probabilityThe probability of each path in the space, and the Bayesian inference model comprises the following steps:the probability of each path is represented, where,according to the maximum probability calculatedAnd planning the path.

Further, planning a path from the start semantic region to the target semantic region based on semantic regions comprises: planning a path from the starting semantic region to the target semantic region includes a predetermined connection path, where the predetermined connection path is preset, and the starting semantic region and the target semantic region are in different predetermined spaces.

Further, planning a path from the start semantic region to the target semantic region based on semantic regions comprises: mapping semantic regions in the predetermined space into semantic identifiers on a map corresponding to the predetermined space, wherein each semantic region comprises one or more semantic identifiers, and the semantic identifiers are feature points of objects in the semantic regions; and planning the path according to one or more semantic identifications corresponding to the semantic area.

Further, planning the path according to one or more semantic identifiers corresponding to the semantic area includes: planning a path from the starting semantic region to the target semantic region according to a particle filter model.

Further, under the condition that the equipment can collect semantic identifiers during moving, according to the semantic identifiers on the map corresponding to the preset space, the adjacent semantic identifiers of the last semantic identifier collected by the equipment are determined, the probability that the equipment collects the semantic identifiers after moving is determined, and according to the semantic identifiers, the equipment collects the semantic identifiers after movingThe particle filter model comprises the following components: n is a radical of_LRepresenting the number of the adjacent semantic tags; n is a radical of_PRepresenting the number of particles in the particle filter model; the semantic identificationIsWherein, theRepresenting a semantic identity l, Q in said predetermined space K^CIndicating that there is only one semantic identifier in said predetermined space C, saidRepresenting a semantic identifier M in said predetermined space M, saidRepresenting a semantic identifier R in the predetermined space R; q. q.s_tRepresenting a current state of the device; o represents the collection of any semantic identifier;a particle i representing a current state;the particle i representing the current state is assigned toThe probability of (d);representing collected semantic tagsThe probability of (d);representing the current weight of the ith particle.

Further, the identifier transition probability is the identifier probability of the current semantic identifier acquired by the device after the last semantic identifier is acquired, wherein the semantic identifier is an adjacent semantic identifier; the tag transition probability is determined based on the separation between two semantic tags associated with the device movement, and the particle filter model comprises: the formula for the label transition probability is interpreted as:wherein,representing semantic tags in the CollectionThen, the semantic mark is collectedThe probability of (a) of (b) being,representing the label transition probability, m and n are the adjacent semantic labels, and K and H are the adjacent semantic regions; the formula of the label transition probability is as follows:where d denotes the separation between two semantic identifiers related to the movement of the device, σ_dRepresents the standard deviation of the interval d;representing semantic tagsAnd semantic identificationThe interval therebetween: wherein,representation in semantic identificationAnd semantic identificationAre adjacent to each other;representing semantic tagsAnd semantic identificationThere is a semantic mark between them;representing semantic tagsAnd semantic identificationThere is a predetermined space with only one semantic identifier in between;representing predetermined spatial and semantic tags with only one semantic tagThe spacing therebetween; wherein,representing predetermined spatial and semantic tags with only one semantic tagAdjacent;representing predetermined spatial and semantic tags with only one semantic tagThere is also a semantic mark in between.

According to another aspect of the embodiments of the present invention, there is also provided a navigation device, including: an acquisition unit, configured to acquire a start semantic region and a target semantic region in a predetermined space, where the predetermined space is divided into a plurality of semantic regions, the plurality of semantic regions include the start semantic region and the target semantic region, and each semantic region includes a feature for identifying the semantic region from the plurality of semantic regions; a planning unit for planning a path from the start semantic area to the target semantic area based on a semantic area; and the moving unit is used for enabling the equipment to move from the starting semantic area to the target semantic area according to the path.

Further, the mobile unit includes: the identification module is used for identifying the obtained image through the equipment to obtain a semantic area corresponding to the image, wherein the semantic area identified by the equipment is used for determining the semantic area where the equipment is located at present; and the moving module is used for moving the equipment to the target semantic area according to the semantic area where the equipment is located at present and the path.

Further, the identification module includes: the first acquisition module is used for acquiring the characteristic points in the image; the matching module is used for matching the feature points in the image with the feature points in a feature point library corresponding to each preset semantic region, wherein one or more feature points corresponding to each semantic region are recorded in the feature point library; a first determining module, configured to determine whether a number of feature points in the image matched with feature points in a feature point library corresponding to each semantic region satisfies a predetermined condition; and the second determining module is used for determining the semantic area corresponding to the image under the condition that the preset condition is met.

Further, the obtaining module comprises: a third determining module, configured to determine whether the number of feature points in the image belongs to a predetermined range; and the second acquisition module is used for acquiring the characteristic points in the image under the condition of belonging to a preset range.

Further, the first determining module comprises: the comparison module compares the matching quantity of the feature points in the image with the feature points in the feature point library corresponding to each semantic region with a preset threshold value; a fourth determining module for determining that the number of matches is greater than a predetermined threshold and satisfies the predetermined condition.

Further, the first determining module comprises: the sorting module is used for sorting each semantic region according to the matching number of the feature points in the image and the feature points in the feature point library corresponding to each semantic region from a large number to a small number; and the fifth determining module is used for determining that the semantic region with the top ranking order meets the preset condition.

Further, the planning unit includes: the analysis module is used for obtaining a semantic relation in the task instruction based on semantic analysis of the task instruction; a sixth determining module, configured to determine the starting semantic area and the target semantic area according to the semantic relationship, and plan a path from the starting semantic area to the target semantic area.

Further, the planning unit includes: the first mapping module is used for mapping the semantic area in the preset space into a regular graph block on a map corresponding to the preset space; and the first planning submodule is used for planning the path according to the rule graph block corresponding to the semantic area.

Further, the first planning sub-module includes: and the first modeling module is used for planning a path from the starting point semantic region to the target semantic region according to a Bayesian inference model.

Further, the planning unit includes: the predetermined module is configured to plan a path from the starting point semantic area to the target semantic area to include a predetermined connection path, where the predetermined connection path is preset, and the starting point semantic area and the target semantic area are in different predetermined spaces.

Further, the planning unit includes: the second mapping module is used for mapping semantic areas in the preset space into semantic identifiers on a map corresponding to the preset space, wherein each semantic area comprises one or more semantic identifiers, and the semantic identifiers comprise the characteristics of the semantic areas; and the second planning submodule is used for planning the path according to one or more semantic identifications corresponding to the semantic area.

Further, the second planning sub-module includes: and the second modeling module is used for planning a path from the starting semantic region to the target semantic region according to a particle filter model.

In the embodiment of the invention, a starting point semantic region and a target semantic region in a predetermined space are obtained, wherein the predetermined space is divided into a plurality of semantic regions, the plurality of semantic regions comprise the starting point semantic region and the target semantic region, and each semantic region comprises a feature for identifying the semantic region from the plurality of semantic regions; planning a path from the starting semantic region to the target semantic region based on a semantic region; and the equipment is enabled to move from the starting point semantic area to the target semantic area according to the path, so that the technical problem that the navigation semantic range is too wide is solved.

Drawings

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention without limiting the invention. In the drawings:

FIG. 1 is a schematic diagram of a navigation method according to an embodiment of the invention;

FIG. 2 is a schematic diagram of an alternative semantic area division according to an embodiment of the present invention;

FIG. 3 is a schematic diagram of an alternative semantic area topology according to an embodiment of the invention;

FIG. 4 is a diagram illustrating an alternative semantic area mapping to a graphical block, according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of an alternative Bayesian network in accordance with embodiments of the present invention;

FIG. 6 is a schematic diagram of an alternative particle filter model in accordance with an embodiment of the present invention;

fig. 7 is a schematic diagram of a navigation device according to an embodiment of the invention.

Detailed Description

In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

In accordance with embodiments of the present invention, there is provided a navigation method and apparatus embodiments, it is noted that the steps illustrated in the flowchart of the drawings may be performed in a computer system such as a set of computer executable instructions, and that while a logical order is illustrated in the flowchart, in some cases the steps illustrated or described may be performed in an order different than here.

Fig. 1 is a schematic diagram of a navigation method according to an embodiment of the present invention, as shown in fig. 1, the method includes the following steps:

step S102, a starting point semantic region and a target semantic region in a predetermined space are obtained, wherein the predetermined space is divided into a plurality of semantic regions, the plurality of semantic regions comprise the starting point semantic region and the target semantic region, and each semantic region comprises a feature for identifying the semantic region from the plurality of semantic regions;

step S104, planning a path from the starting semantic region to the target semantic region based on the semantic region;

step S106, the equipment is moved from the starting semantic area to the target semantic area according to the path.

According to the above-described embodiment of the present invention, in the predetermined space, the predetermined space is divided into the plurality of semantic regions according to the plurality of features within the predetermined space, each semantic region is made to include a feature for identifying the semantic region from the plurality of semantic regions, and the divided semantic regions include the start semantic region and the target semantic region of the device, and the device can move from the start semantic region to the target semantic region along the planned path according to the planned path from the start semantic region to the target semantic region. The preset space is divided into a plurality of semantic areas according to the characteristics, and the moving path of the equipment is planned according to the divided semantic areas, so that the equipment can accurately navigate from the starting point semantic area to the target semantic area according to the planned path, and the technical problem that the navigation semantic range is too wide is solved.

In the above embodiment, the predetermined space for dividing the semantic region may be a space in which the device works or moves, the determination of the space may be set manually by a user of the device or may be identified according to a preset program of the device, the space in which the device works or moves may be divided into one or more predetermined spaces, and one or more predetermined spaces may be subjected to semantic region division.

In the above embodiment, the semantic area division on the predetermined space may be implemented according to a division rule preset by the device, or may be implemented by artificially setting the division rule. As an optional embodiment, in a certain predetermined space, the device user may divide the predetermined space into a plurality of semantic regions by a manual setting manner, where the division of the semantic regions may be performed according to the functions of the semantic regions, or may be performed according to objects in the semantic regions, and the device may determine the semantic regions by recognizing the functions of the semantic regions, or recognizing the objects in the semantic regions.

There may be an area in the predetermined space that is blocked by an object or is not allowed to be accessed by the device, and in the process of semantically dividing the portion, the area may be divided into a blank semantic area, and the blank semantic area is not accessed by the device after being divided, that is, the device does not pass through the semantic area. By dividing the blank semantic area, certain areas which are not allowed to be accessed by the device can be isolated from the device.

Alternatively, the device performing the activity in the predetermined space may be a robot device.

Fig. 2 is a schematic diagram of an optional semantic area division according to an embodiment of the present invention, as shown in fig. 2, the device moves in a bedroom, a hall, and a kitchen, and divides each room into a plurality of semantic areas, wherein a shaded area where each roman numeral is located represents one semantic area, and an area labeled with a shade and a roman letter is a blank semantic area; the connecting lines of adjacent hexagons represent the boundary of each semantic region, and each hexagons used for representing the boundary of each semantic region is marked by English letters.

In the division of the semantic regions in the predetermined space, the room can be divided into a plurality of nameable and identifiable semantic regions according to some objects with specific functions or appearances in the room. For example, there are refrigerators, microwave ovens, gas ranges, etc. in the kitchen, beds, televisions, etc. in the bedroom, and each room may be divided into a plurality of semantic regions according to the above-mentioned features in the room.

The semantic areas may be divided according to the function of the area, for example, a kitchen may be divided into a food storage area and a processing area, a restaurant may be divided into a dining area, and a bedroom may be divided into a resting area and a sleeping area. The semantic areas may also be divided according to the appearance of objects in the area, for example, a kitchen may be divided into a refrigerator area, a microwave oven area, a bedroom may be divided into a television area, a sofa area, a bed area, and so on. In the process of carrying out semantic region division on a room, a division rule of a semantic region can be set artificially according to the requirement of a user.

And planning a path from the starting point semantic area to the target semantic area by the equipment based on the divided semantic areas, so that the equipment moves from the starting point semantic area to the target semantic area according to the path. In order to ensure that the equipment can accurately move according to the planned path, the semantic area where the equipment is located needs to be judged constantly, and the moving mode is adjusted. As an alternative embodiment, causing the device to move from the start semantic area to the target semantic area according to the path includes: identifying the obtained image through equipment to obtain a semantic area corresponding to the image, wherein the semantic area identified by the equipment is used for determining the semantic area where the equipment is located at present; and moving the equipment to the target semantic area according to the current semantic area and the path of the equipment. The equipment acquires the image information of the area where the equipment is located in real time through the image acquisition module, determines the semantic area where the equipment is located at present according to the acquired image information, and can judge whether the equipment moves to the target semantic area according to the planned path, so that the equipment can move in the specified semantic area route according to the planned route.

In order to enable the manner of determining the semantic area where the device is currently located in the foregoing embodiment to be performed by a computer, as an optional embodiment, the obtaining, by the device, the semantic area corresponding to the image by identifying the obtained image includes: acquiring feature points in an image; matching the feature points in the image with the feature points in a feature point library corresponding to each preset semantic region, wherein one or more feature points corresponding to each semantic region are recorded in the feature point library; determining whether the matching number of the feature points in the image and the feature points in the feature point library corresponding to each semantic region meets a preset condition; and determining a semantic area corresponding to the image under the condition that a preset condition is met.

Specifically, feature points of an image obtained by equipment are extracted, the extracted feature points are matched with feature points in a feature point library corresponding to each semantic region, the number of the feature points in the image matched with the feature points in the feature point library corresponding to each semantic region is recorded according to a matching result, and whether the number of the feature point matches meets a preset condition or not is judged to determine the equipmentThe semantic region, wherein one or more feature points of the semantic region are recorded in a feature point library corresponding to each semantic region. For the above process, it can be formulated: n is a radical of_n,i＝I_n∩Q_i1,2, 1, 3, N, where N is N_n,iRepresenting the nth image I in the image_nAnd the ith regional feature point library Q_iAnd matching the feature points to obtain a matched feature point set. According to the scheme, the characteristic points in the image are matched with the characteristic points in the preset characteristic point library, so that the mode of determining the semantic area corresponding to the image can be realized through calculation, and further, the equipment can automatically identify the semantic area where the equipment is located according to the method.

Alternatively, the feature points used for matching in the feature point library corresponding to each semantic region may be all feature points in the feature point library corresponding to the semantic region.

In the process of identifying the semantic area where the image is located, the device needs to match the feature points in the image with the feature points in the feature point library corresponding to each semantic area, wherein as an optional embodiment, the feature point library corresponding to each semantic area may be established in a manner of acquiring the image. The feature point library corresponding to each semantic area is equivalent to that of equipment in the semantic area, all feature points in the image are extracted, and then the corresponding feature point library can be established.

In the process of identifying the semantic area where the image is located, because the obtained image is randomly obtained, some images may have the situation that the number of feature points in the image is too small or too large. Determining the semantic area where the image is located according to the too few number of the feature points in the image, wherein the semantic area is inaccurate; in the process of determining the semantic area where the equipment is located according to the image with the excessive number of the feature points in the image, the matching process is slow due to the fact that the number of the feature points is large. In order to solve the above problem, the present application provides an alternative embodiment, which acquires feature points in an image, and includes: determining whether the number of feature points in the image belongs to a predetermined range; in the case of belonging to a predetermined range, a feature point in the image is acquired. According to the scheme, the characteristic points in the image are matched with the preset range, the image with the number of the characteristic points belonging to the preset range is determined, and the characteristic points in the image are obtained.

As an alternative embodiment, in the case where the number of feature points in an image is small, the number of feature points that can be referred to in the image is small, the number of feature points that can be referred to is easily changed, and the error in determining the semantic region where the image is located with the feature points is high, so that matching is discarded for the image. For the judgment of the images, the feature points in the images can be matched with a preset range, and the images with the number of the feature points lower than the preset range are determined to be the images which abandon the matching.

As another alternative, in the case that the number of feature points in an image is large, for example, a wall where wallpaper with colored speckles all regularly distributed is pasted on one side, when the image is acquired, the number of feature points that is hundreds times larger than that of a common image may be detected, which may interfere with the establishment of a feature point library, and may also significantly slow down the matching speed. For the judgment of the images, the characteristic points in the images can be matched with the preset range, and the images with the characteristic points higher than the preset range are determined to be the images which are abandoned for comparison.

In the process of determining whether the number of matched feature points meets the predetermined condition, as an alternative embodiment, determining whether the number of matches between feature points in the image and feature points in the feature point library meets the predetermined condition includes: comparing the matching number of the feature points in the image and the feature points in the feature point library corresponding to each semantic region with a preset threshold value; it is determined that the number of matches is greater than a predetermined threshold satisfying a predetermined condition. In this embodiment, the number of matched feature points is obtained by matching the feature points in the image with the feature points in the feature point library corresponding to each semantic region, and the image is directly determined to be the image satisfying the predetermined condition when the number of matched feature points is greater than the threshold value.

In the process of determining whether the number of feature point matches meets the predetermined condition, as an optional embodiment, determining whether the number of feature point matches between feature points in the image and feature points in the feature point library meets the predetermined condition includes: according to the matching number of the feature points in the image and the feature points in the feature point library corresponding to each semantic region, sequencing each semantic region according to the sequence of the matching number from more to less; and determining that the semantic region with the top ranking order meets a preset condition. The semantic region with the top ranking order represents that the number of the feature points in the feature point library corresponding to the semantic region is the largest compared with the feature points in the image, the closest semantic region of the image can be determined, and the semantic region where the image is located can be accurately determined.

In order to more accurately determine the semantic region where the image is located through the predetermined condition, as an optional embodiment, planning a path from the starting semantic region to the target semantic region based on the semantic region includes: obtaining a semantic relation in the task instruction based on semantic analysis of the task instruction; and determining a starting semantic region and a target semantic region according to the semantic relation, and planning a path from the starting semantic region to the target semantic region. The equipment carries out semantic analysis on a task instruction sent by a user to obtain one or more semantic regions including a target semantic region in the task instruction, and plans a path from a starting point semantic region to the target semantic region according to the relation between the semantic regions.

Fig. 3 is a schematic diagram of an alternative semantic area topology according to an embodiment of the present invention, as shown in fig. 3.

The task instruction includes one or more semantic regions, the relationship between the semantic regions may be identified by a topology structure, the schematic diagram of semantic region division shown in fig. 2 is mapped to a corresponding topology structure, as shown in fig. 3, the semantic regions are a mesh structure, and the semantic regions are connected by the topology structure, where each circle represents a semantic region and is represented by a greek number. The equipment analyzes the task instructions according to the received task instructions, determines a topological structure of the relation between the task instructions, and navigates the equipment to a required position according to the topological structure. For example, the device receives a task instruction of "take out the meal from the refrigerator and heat it, and send it to the front of the bed", in which case, the device first needs to analyze the characteristics thereof according to the task instruction, determine the semantic area where the task is located, and plan the path from the device to the semantic area. After the equipment receives the task instruction, the following operations are required to be completed, the equipment moves to a semantic area VII in a kitchen, food is taken from a refrigerator and heated by a microwave oven, the food enters a hall through the semantic areas VI, VII, IV, II and I in the kitchen, and then the food passes through a plurality of semantic areas I in the hall and then reaches a semantic area II where a bed is located through the semantic area I in a bedroom, and the heated food is delivered to the front of the bed. According to the process, a path can be planned. In addition, in the process, it can be found that a single semantic area may include multiple features, for example, the semantic area VII in the kitchen includes three features, namely a refrigerator, a gas cooker and a microwave oven, and the robot completes two tasks of "taking out food from the refrigerator" and "heating" in the semantic area VII in the kitchen.

In the process of planning a path, some semantic areas in a predetermined space are required to be passed by the device, for example, when the device needs to leave a certain room, the device needs to pass through the semantic area where the porch is located to reach the semantic area where the porch is located in the next predetermined space, so that when the area where the device moves includes a plurality of predetermined spaces, the path of the device in each predetermined space (i.e., each room) can be planned first, and then the paths are spliced to obtain a complete planned path. Through the embodiment, the planned path can be divided into the planned paths spliced by the planned paths in the plurality of independent preset spaces, wherein the preset space through which the planned path passes is selected, and then the planned path in the preset space is determined, so that the number of semantic areas required to be referred by the planned path can be reduced, and the process of planning the path is simpler and faster.

Before planning a route from the starting semantic area to the target semantic area, the correspondence of the semantic area in the predetermined space on the map may be established so that the device plans a navigation route on the map according to the rules of the semantic area. As an optional embodiment, mapping a semantic area in a predetermined space into a regular graph block on a map corresponding to the predetermined space; and planning a path according to the rule graph block corresponding to the semantic area. The device is enabled to plan a path on the image block of the semantic area on the corresponding map according to the navigation of the semantic area. The semantic area is mapped into the regular graph blocks on the map corresponding to the preset space, and the equipment moves between the semantic areas, namely the equipment can be considered to move between the regular graph blocks, so that the equipment can correspond the starting point semantic area and the target semantic area to two regular graph blocks on the map, and further the equipment can plan a planned path which can help the equipment to accurately navigate according to the positions of the two regular graph blocks on the map and the distance between the two regular graph blocks.

Fig. 4 is a schematic diagram of mapping an optional semantic area into a graphic block according to an embodiment of the present invention, and as shown in fig. 4, the schematic diagram of dividing the semantic area shown in fig. 2 is mapped into a corresponding regular graphic block on a map, where the roman alphabet in each regular graphic block in fig. 4 is the semantic area corresponding to the roman alphabet in fig. 2, and nul in fig. 4 represents a blank semantic area corresponding to the regular graphic block.

The semantic area in the predetermined space is mapped to the regular pattern block on the map corresponding to the predetermined space, that is, the semantic area of the room where the device is located is mapped to the regular pattern block on the map corresponding to the room, so that the device can determine the moving direction of the device according to the regular pattern block on the map, and the device can be conveniently navigated.

In the process of planning the moving path of the device, as an optional embodiment, planning the path according to the rule graph block corresponding to the semantic area includes: and planning a path from the starting point semantic region to the target semantic region according to a Bayesian inference model. The Bayes method can rapidly calculate the maximum posterior probability by using the Bayes network under the condition of given input, so that the Bayes network can be selected to deduce the path which can best meet the constraint condition.

The Bayes method can rapidly calculate the maximum posterior probability by using the Bayes network under the condition of given input, so that the Bayes network is selected to deduce the path which can most meet the constraint condition.

Fig. 5 is a schematic diagram of an alternative bayesian network according to an embodiment of the present invention, and as shown in fig. 5, a bayesian network is dynamically established according to a start point and an end point of a given path, and a path satisfying a constraint condition is determined.

In the case where the rule graph block is a rectangle and the device can move in the rectangle at least in four directions, namely up, down, left and right, the model of bayesian inference is as follows: o denotes the four directions of the rectangle,the presentation device is able to move in an upward direction of the semantic area N in the predetermined space R,the presentation device is able to move in a downward direction of the semantic area N in the predetermined space R,the presentation device is able to move in the left direction of the semantic area N in the predetermined space R,the presentation device is capable of moving in the right direction of the semantic area N in the predetermined space R.

In the case that the regular image block is a rectangle and the device can determine the number of rotations of the device from a certain semantic area to a target semantic area in the rectangle under the condition that the rotation direction is 90 ° each time, wherein the number of rotations is determined according to the last movement direction of the device to the last semantic area and the current movement direction of the device to the current semantic area, the bayesian inference model comprises the following steps:number of turns representing movement of the device from a previous semantic area to a current semantic area, o_iRepresenting the last direction of movement of the device to the last semantic area, o_jCurrent moving direction, o, representing the device moving to the current semantic area_iAnd o_jThe four moving directions of the upper, the lower, the left and the right can be respectively taken; if it is the previous moving direction o_iAnd the current moving direction o_jSame number of rotationsIf it is the previous moving direction o_iAnd the current moving direction o_j90 deg. difference, the number of rotationsIf it is the previous moving direction o_iAnd the current moving direction o_j180 DEG difference, the number of rotations

Planning the movement path of the device also requires introducing the concept of transition probability. The transition probability is the probability that the equipment moves to the current moving direction of the current semantic region after moving to the previous moving direction of the previous semantic region, wherein the previous semantic region and the current semantic region are adjacent semantic regions; the transition probability is calculated according to the rotation times of the equipment moving from the previous semantic area to the current semantic area and all the rotation directions of the equipment, and the Bayesian inference model comprises the following steps: the formula for the label transition probability is interpreted as:wherein,representing the probability that the device moves to the current direction of movement of the current semantic region after moving to the previous direction of movement of the previous semantic region, wherein,representing the transition probability, R_MAnd R_NAre two adjacent semantic regions in a predetermined space; the calculation formula of the transition probability is as follows:wherein omega_oIndicating the overall rotational direction of the device.

Obtaining the probability of each path in the same preset space according to the recurrence relation and the transition probability of the Bayesian network, wherein the Bayesian inference model comprises the following steps:represents the probability of each path, where M,According to the maximum probability calculatedAnd planning a path.

During the movement of the device, there is a need to pass through some predetermined space, i.e. no action other than that related to the movement is performed within the passed predetermined space. During the process that the equipment passes through the preset space, as an optional embodiment, planning a path from the starting semantic area to the target semantic area based on the semantic areas comprises: planning a path from the starting semantic area to the target semantic area includes a predetermined connection path, wherein the predetermined connection path is preset, and the starting semantic area and the target semantic area are in different predetermined spaces.

Alternatively, in the case where a path from the start semantic region to the target semantic region needs to pass through one or more predetermined spaces, a connection path between the plurality of predetermined spaces is set in advance. The device can move according to the preset connection path among the plurality of preset spaces in the process of passing through some preset spaces, the device does not need to do any work irrelevant to the movement in the moving process, the device can be ensured to rapidly pass through the preset spaces, and the preset connection path can also be the shortest path of the device passing through the space, and the speed of the device passing through the preset spaces can also be increased.

In the process of planning the path, the path may also be planned according to features in a predetermined space, and as an optional embodiment, planning the path from the starting semantic region to the target semantic region based on the semantic regions includes: the semantic area in the preset space comprises one or more semantic identifications corresponding to the semantic area, wherein the semantic identification comprises the characteristics of the semantic area; and planning the path according to one or more semantic identifications corresponding to the semantic area.

Optionally, the planning the path according to one or more semantic identifiers corresponding to the semantic area includes: and planning a path from the starting semantic region to the target semantic region according to the particle filter model.

Fig. 6 is a schematic diagram of an alternative particle filter model according to an embodiment of the invention, as shown in fig. 6.

The hexagons represent semantic labels and the small circles represent particles. The hexagons in the major axis ellipse represent the semantic identifiers that are actually acquired, and the hexagons in the minor axis ellipse represent the semantic identifiers that are acquired with a high probability. The distance of a small circle from all hexagons represents the probability that this particle is assigned to all semantic labels, and the probability is inversely proportional to the distance. The model state is the probability of collecting the semantic identifier, wherein the probability of the semantic identifier is the weight of the semantic identifier.

Determining the probability of the current semantic identifier acquired by the equipment according to the adjacent semantic identifier where the equipment is located and the last semantic identifier acquired by the equipment, wherein the method comprises the following steps according to a particle filter model: n is a radical of_LRepresenting the number of semantic tags; n is a radical of_PRepresents the number of particles;representing a set of statesWherein,semantic identifier l, Q representing a predetermined space K_CIndicating that there is only one semantic identifier in the predetermined space C,Q_C；q_tindicating the state of the device at time t; o represents the collection of any semantic identifier;representing the state of the particle i at time t;indicating that particle c is assigned to at time tThe probability of (d);representing acquisition semantic tagsThe probability of (d);representing the weight of the ith particle at time t.

Determining the identifier transfer probability of the equipment according to the probability of the current semantic identifier acquired after the last semantic identifier is acquired by the equipment, wherein the following steps are included according to a particle filter model:indicating when the last semantic mark was collectedLater collecting current semantic markThe probability of (a), wherein,representing label transition probability, m and n are adjacent semantic labels, and K and H are adjacent semantic regions; the formula for the label transition probability is:wherein d represents the interval between two related semantic identifiers; sigma_dRepresents the standard deviation of d;representation in semantic identificationAndthere is no other semantic identification in between;representing semantic tagsAndthere is a semantic mark between them;the two semantic identifiers are in different predetermined spaces, and the two predetermined spaces are connected by the predetermined space with only one semantic identifier;representing predetermined spatial and semantic tags with only one semantic tagThe spacing therebetween;representing semantic tagsAdjacent to a predetermined space having only one semantic identifier;representing a predetermined space having only one semantic identifierThere is also a semantic mark in between.

Assuming that each semantic tag is independently collected, the state vector is composed of the probabilities of the collected semantic tags, and the following is included according to the particle filter model:representing a state vector.

In the case where no new semantic tags are collected, the state vector is invariant, comprising according to the particle filter model:wherein, w_tA vector representing the composition of white noise,

obtaining an observation model of each semantic identifier according to the state vector, wherein the observation model comprises the following components according to a particle filter model:

for each semantic tag, the observation model can be represented as:

wherein v is_tA vector representing the composition of white noise, representing that the semantic mark m is collected in a predetermined space K; z_tRepresents the probability of observing the semantic representation, and Z_tIndicating the specific gravity of each semantic identification collected.

According to the actually acquired semantic identifiers, calculating the acquired probability of all the semantic identifiers through a total probability formula, wherein the particle filter model comprises the following steps:is a total probability formula.

All particles in the particle filter model are distributed to each semantic identifier according to random probability, all semantic regions are collected according to equal probability, and the method comprises the following steps according to the particle filter model:represents t₀Temporal collection of semantic tagsThe probability of (c).

The weight collected to each semantic identifier is represented by a probability, and the probability space description comprises the following steps: weight of each particleThe assignment to each semantic tag may be expressed as:wherein,w(cⁱ) Expressed as a weight for each particle, the weight can be calculated by the distance of the probability space:

standardizing the weight of each particle according to the weight of each particle to ensure that the sum of all particles is 1, calculating the probability of each particle distributed to a certain semantic identifier according to the weight, updating the probability of all semantic identifiers, planning a path according to the probability of all semantic identifiers, and according to a particle filter model, the method comprises the following steps:a normalization formula representing a weight of the particles; the particles realize resampling according to the weight; according to the weight w (c)ⁱ) Computing each particle assigned to a semantic identifierProbability of (2)Updating the probability of the full semantic identificationAnd planning a path according to the maximum probability of all semantic identifiers, wherein the probability formula of all semantic identifiers is as follows:

fig. 7 is a schematic diagram of a navigation device according to an embodiment of the present invention, as shown in fig. 7, the device including: an acquisition unit 71 configured to acquire a start point semantic region and a target semantic region in a predetermined space, where the predetermined space is divided into a plurality of semantic regions, the plurality of semantic regions including the start point semantic region and the target semantic region, each of the semantic regions including a feature for identifying the semantic region from the plurality of semantic regions; a planning unit 72 for planning a path from the start semantic region to the target semantic region based on the semantic regions; a moving unit 73 for causing the device to move from the start semantic area to the target semantic area according to the path.

As an alternative embodiment, the mobile unit comprises: the recognition module is used for recognizing the obtained image through the equipment to obtain a semantic area corresponding to the image, wherein the semantic area recognized by the equipment is used for determining the semantic area where the equipment is located at present; and the moving module is used for moving the equipment to the target semantic area according to the semantic area and the path where the equipment is located at present.

As an alternative embodiment, the identification module comprises: the first acquisition module is used for acquiring feature points in the image; the comparison module is used for comparing the feature points in the image with the feature points in a feature point library corresponding to each preset semantic region, wherein one or more feature points corresponding to each semantic region are recorded in the feature point library; the first determining module is used for determining whether the matching number of the feature points in the image and the feature points in the feature point library corresponding to each semantic region meets a preset condition or not; and the second determining module is used for determining the semantic area corresponding to the image under the condition that the predetermined condition is met.

As an alternative embodiment, the obtaining module includes: a third determining module, configured to determine whether the number of feature points in the image belongs to a predetermined range; and the second acquisition module is used for acquiring the characteristic points in the image under the condition of belonging to the preset range.

As an alternative embodiment, the first determining module comprises: the matching module compares the matching quantity of the feature points in the image and the feature points in the feature point library corresponding to each semantic region with a preset threshold value; and the fourth determination module is used for determining that the number of the matches is larger than the preset threshold value and meets the preset condition.

As an alternative embodiment, the first determining module comprises: the sequencing module is used for sequencing each semantic region according to the matching number of the feature points in the image and the feature points in the feature point library corresponding to each semantic region from high to low; and the fifth determining module is used for determining that the semantic region with the top ranking order meets the preset condition.

As an alternative embodiment, the planning unit comprises: the analysis module is used for obtaining semantic relations in the task instructions based on semantic analysis of the task instructions; and the sixth determining module is used for determining the starting point semantic area and the target semantic area according to the semantic relation and planning a path from the starting point semantic area to the target semantic area.

As an alternative embodiment, the planning unit comprises: the first mapping module is used for mapping the semantic area in the preset space into a regular graph block on a map corresponding to the preset space; and the first planning submodule is used for planning a path according to the rule graphic block corresponding to the semantic area.

As an alternative embodiment, the first planning submodule includes: and the first modeling module is used for planning a path from the starting point semantic region to the target semantic region according to a Bayesian inference model.

As an alternative embodiment, the planning unit comprises: the system comprises a predetermined module, a target semantic region and a processing module, wherein the predetermined module is used for planning a path from the starting point semantic region to the target semantic region to comprise a predetermined connection path, the predetermined connection path is preset, and the starting point semantic region and the target semantic region are in different predetermined spaces.

As an alternative embodiment, the planning unit comprises: the second mapping module is used for mapping semantic areas in the preset space into semantic identifiers on a map corresponding to the preset space, wherein each semantic area comprises one or more semantic identifiers, and each semantic identifier comprises the characteristics of the semantic area; and the second planning submodule is used for marking a planning path according to one or more semantics corresponding to the semantic area.

As an alternative embodiment, the second planning submodule includes: and the second modeling module is used for planning a path from the starting point semantic region to the target semantic region according to the particle filter model.

The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.

In the above embodiments of the present invention, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.

In the embodiments provided in the present application, it should be understood that the disclosed technology can be implemented in other ways. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units may be a logical division, and in actual implementation, there may be another division, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, units or modules, and may be in an electrical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic or optical disk, and other various media capable of storing program codes.

The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and decorations can be made without departing from the principle of the present invention, and these modifications and decorations should also be regarded as the protection scope of the present invention.

Claims

1. A navigation method, comprising:

acquiring a starting point semantic region and a target semantic region in a predetermined space, wherein the predetermined space is divided into a plurality of semantic regions, the plurality of semantic regions comprise the starting point semantic region and the target semantic region, and each semantic region comprises a feature for identifying the semantic region from the plurality of semantic regions;

planning a path from the starting semantic region to the target semantic region based on a semantic region;

causing a device to move from the start semantic area to the target semantic area according to the path.

2. The method of claim 1, wherein causing the device to move from the origin semantic region to the target semantic region according to the path comprises:

identifying the obtained image through the equipment to obtain a semantic area corresponding to the image, wherein the semantic area identified by the equipment is used for determining the semantic area where the equipment is located at present;

and moving the equipment to the target semantic area according to the semantic area where the equipment is located at present and the path.

3. The method of claim 2, wherein identifying, by the device, the obtained image to obtain the semantic region corresponding to the image comprises:

acquiring feature points in the image;

matching the feature points in the image with the feature points in a feature point library corresponding to each preset semantic region, wherein one or more feature points corresponding to each semantic region are recorded in the feature point library;

determining whether the matching number of the feature points in the image and the feature points in the feature point library corresponding to each semantic region meets a preset condition;

and determining a semantic area corresponding to the image under the condition that the preset condition is met.

4. The method of claim 3, wherein obtaining feature points in the image comprises:

determining whether the number of feature points in the image belongs to a predetermined range;

and acquiring the characteristic point in the image under the condition of belonging to a preset range.

5. The method of claim 3, wherein determining whether the number of feature points in the image that match feature points in the feature point library satisfies a predetermined condition comprises:

comparing the matching quantity of the feature points in the image with the feature points in the feature point library corresponding to each semantic region with a preset threshold value;

determining that the number of matches is greater than a predetermined threshold meeting the predetermined condition.

6. The method of claim 3, wherein determining whether the number of feature points in the image that match feature points in the feature point library satisfies a predetermined condition comprises:

according to the matching number of the feature points in the image and the feature points in the feature point library corresponding to each semantic region, sequencing each semantic region according to the sequence of the matching number from high to low;

determining that the semantic region with the top ranking order meets a preset condition.

7. The method of claim 1, wherein planning a path from the start semantic region to the target semantic region based on semantic regions comprises:

obtaining a semantic relation in a task instruction based on semantic analysis of the task instruction;

and determining the starting point semantic region and the target semantic region according to the semantic relation, and planning a path from the starting point semantic region to the target semantic region.

8. The method of claim 1, wherein planning a path from the start semantic region to the target semantic region based on semantic regions comprises:

mapping the semantic area in the preset space into a regular graph block on a map corresponding to the preset space;

and planning the path according to the rule graphic block corresponding to the semantic area.

9. The method of claim 8, wherein planning the path according to the rule-graph block corresponding to the semantic area comprises:

planning a path from the starting point semantic region to the target semantic region according to a Bayesian inference model.

10. The method according to claim 9, wherein in case the rule graph block is a rectangle and the device can only move in four directions, up, down, left and right, in the rectangle, the model of bayesian inference is as follows:

o denotes the four directions of the rectangle,indicating that the device is capable of moving in an upward direction of the semantic region N in the predetermined space R,indicating that the device is capable of moving in a downward direction of the semantic region N in the predetermined space R,indicating that the device is capable of moving in the leftward direction of the semantic region N in the predetermined space R,indicating that the device is capable of moving in a rightward direction of the semantic region N in the predetermined space R.

11. The method according to claim 10, wherein the rule graph block is a rectangle, and the device is capable of determining the number of rotations of the device from a certain semantic area to a target semantic area in the rectangle under the condition that each rotation direction is 90 °, wherein the number of rotations is determined according to the last movement direction of the device to the last semantic area and the current movement direction of the device to the current semantic area, and the bayesian inference model comprises the following steps:

representing the number of turns of the device to move from a previous semantic area to a current semantic area, o_iRepresents the last direction of movement of the device, the o_jRepresents the current direction of movement of the device, the o_iAnd said o_jThe four moving directions of the upper, the lower, the left and the right can be respectively taken;

if the last moving direction o_iAnd the current moving direction o_jSame, then the number of rotations

If the last moving direction o_iAnd the current moving direction o_jBy 90 deg., the number of rotations

If the last moving direction o_iAnd the current moving direction o_jBy 180 deg., the number of rotations

12. The method of claim 11, wherein transition probability is a probability that the device moves to a current moving direction of a current semantic region after moving to a previous moving direction of a previous semantic region, wherein the previous semantic region and the current semantic region are adjacent semantic regions; the transition probability is calculated according to the number of turns of the device moving from the previous semantic region to the current semantic region and all rotation directions of the device, and the Bayesian inference model comprises the following steps:

the formula for the transition probability is interpreted as:wherein,representing a probability that the device moves to a current movement direction of a current semantic region again after moving to a previous movement direction of the previous semantic region, wherein,representing the transition probability, R_MAnd R_NAre two adjacent semantic regions in the predetermined space;

the calculation formula of the transition probability is as follows:wherein omega_oIndicating the overall rotational direction of the device.

13. The method according to claim 12, wherein the probability of each path in the same predetermined space is obtained according to the recurrence relation of the bayesian network and the transition probability, and the bayesian inference model comprises:

the probability of each path is represented, where,according to the most calculatedHigh probabilityAnd planning the path.

14. The method of claim 1, wherein planning a path from the start semantic region to the target semantic region based on semantic regions comprises:

planning a path from the starting semantic region to the target semantic region includes a predetermined connection path, where the predetermined connection path is preset, and the starting semantic region and the target semantic region are in different predetermined spaces.

15. The method of claim 1, wherein planning a path from the start semantic region to the target semantic region based on semantic regions comprises:

mapping semantic regions in the predetermined space into semantic identifiers on a map corresponding to the predetermined space, wherein each semantic region comprises one or more semantic identifiers, and the semantic identifiers are feature points of objects in the semantic regions;

and planning the path according to one or more semantic identifications corresponding to the semantic area.

16. The method of claim 15, wherein planning the path according to one or more semantic tags corresponding to the semantic region comprises:

planning a path from the starting semantic region to the target semantic region according to a particle filter model.

17. The method according to claim 16, wherein, in a case where the device is capable of acquiring semantic tags during movement, determining semantic tags adjacent to the last semantic tag acquired by the device according to the semantic tags on the map corresponding to the predetermined space, and determining the probability that the device acquires the semantic tags after movement, the following steps are included according to the particle filter model:

N_Lrepresenting the number of the adjacent semantic tags; n is a radical of_PRepresenting the number of particles in the particle filter model; the semantic identificationIsWherein, theRepresenting a semantic identity l, Q in said predetermined space K^CIndicating that there is only one semantic identifier in said predetermined space C, saidRepresenting a semantic identifier M in said predetermined space M, saidRepresenting a semantic identifier R in the predetermined space R; q. q.s_tRepresenting a current state of the device; o represents the collection of any semantic identifier;a particle i representing a current state;the particle i representing the current state is assigned toThe probability of (d);presentation collectionTo semantic identificationThe probability of (d);representing the current weight of the ith particle.

18. The method of claim 17, wherein the token transition probability is a token probability that the device moves to acquire a current token after acquiring a last token, wherein the token is an adjacent token; the tag transition probability is determined based on the separation between two semantic tags associated with the device movement, and the particle filter model comprises:

the formula for the label transition probability is interpreted as:wherein,representing semantic tags in the CollectionThen, the semantic mark is collectedThe probability of (a) of (b) being,representing the label transition probability, m and n are the adjacent semantic labels, and K and H are the adjacent semantic regions;

the formula of the label transition probability is as follows:wherein,d represents the interval between two semantic tags related to the movement of the device, σ_dRepresents the standard deviation of the interval d;

representing semantic tagsAnd semantic identificationThe interval therebetween: wherein,

representation in semantic identificationAnd semantic identificationAre adjacent to each other;

representing semantic tagsAnd semantic identificationThere is a semantic mark between them;

representing semantic tagsAnd semantic identificationThere is a predetermined space with only one semantic identifier in between;

representing predetermined spatial and semantic tags with only one semantic tagThe spacing therebetween; wherein,

representing predetermined spatial and semantic tags with only one semantic tagAdjacent;

representing predetermined spatial and semantic tags with only one semantic tagThere is also a semantic mark in between.

19. The method of claim 18, wherein the state vector of probabilities is composed of probabilities of semantic identifications collected by the device during movement, and wherein the particle filter model comprises:

representing the state vector.

20. The method of claim 19, wherein the state vector is invariant without new semantic tags being collected, the state transition models for all particle filters according to which comprise:

wherein, w_tA vector representing the composition of white noise,

21. the method of claim 20, wherein the obtaining of the observation model for each semantic identifier in the predetermined space is based on state transition models of all particle filters, and wherein the obtaining of the observation model for each semantic identifier in the predetermined space based on the particle filter models comprises:

for each semantic tag, the observation model can be represented as:

wherein v is_tA vector representing the composition of white noise,

representing that the semantic mark m is collected in a predetermined space K;

Z_trepresenting the probability of observing the semantic mark by an observation model, and Z_tIndicating the specific gravity of each semantic identification collected.

22. The method of claim 21, wherein extracting a full probability formula in the observation model according to each semantically identified observation model in the predetermined space comprises:

is a total probability formula.

23. The method of claim 22, wherein calculating the probability that all semantic tags are collected based on the full probability formula and the actually collected semantic tags comprises assigning all particles in a particle filter model to each semantic tag according to a random probability, and deriving the probability that each semantic tag is collected based on the number of particles and the number of all ions assigned to each semantic tag, wherein the following is included in the particle filter model:

representing collected semantic tagsThe probability of (c).

24. The method of claim 23, wherein the weight of each semantic identifier collected is represented by a probability, and wherein the probability space description comprises:

weight of each particleThe assignment to each semantic tag may be expressed as:

wherein,w(cⁱ) The weight, expressed as a weight for each particle, can be calculated by the distance of the probability space:

25. the method of claim 24, wherein the weight of each particle is normalized according to the weight of each particle to ensure that the sum of all particles is 1, the probability of each particle being assigned to a semantic tag is calculated according to the weight, the probability of all semantic tags is updated, and a path is planned according to the probability of all semantic tags, and the particle filter model comprises:

a normalization formula representing a weight of the particles;

the particles realize resampling according to the weight;

according to the weight w (c)ⁱ) Computing each particle assigned to a semantic identifierProbability of (2)

Updating the probability of the full semantic identificationAnd planning the path according to the maximum probability of all semantic identifiers, wherein the probability formula of all semantic identifiers is as follows:

26. a navigation device, comprising:

an acquisition unit, configured to acquire a start semantic region and a target semantic region in a predetermined space, where the predetermined space is divided into a plurality of semantic regions, the plurality of semantic regions include the start semantic region and the target semantic region, and each semantic region includes a feature for identifying the semantic region from the plurality of semantic regions;

a planning unit for planning a path from the start semantic area to the target semantic area based on a semantic area;

and the moving unit is used for enabling the equipment to move from the starting semantic area to the target semantic area according to the path.