CN113163484B - Indoor positioning method - Google Patents

Abstract

The invention discloses an indoor positioning method, which comprises the following steps: in an off-line stage, according to the indoor structure and layout, performing area division on the indoor space, establishing corresponding coordinate axes, setting reference points in the area, and acquiring RSSI (received signal strength indicator) of wireless signals at the reference points by using different types of equipment to form fingerprint vectors so as to construct a fingerprint database; the BP neural network fits the RSSI of the wireless signal; constructing an electronic map, and storing information into a database; in the on-line positioning stage, firstly, carrying out area positioning and determining an area of a point to be positioned; then, positioning the position in the area, and determining the position coordinates of the position to be positioned; determining a reference point set which needs to pass by to reach the target position according to the target information, and then determining whether the current position advances to the target position or not by combining the current position information; and updating the fingerprint database in real time according to various feedback information. The invention can update the fingerprint database in time, enhance the real-time performance of the fingerprint database and have high positioning precision.

Description

An Indoor Positioning Method

technical field

The invention relates to a positioning method, in particular to an indoor positioning method.

Background technique

Due to the complex indoor environment and natural isolation reasons such as wall occlusion, satellite signals are severely attenuated indoors, which cannot meet the accuracy requirements of indoor positioning or even complete indoor positioning.

The traditional ultrasonic and infrared positioning technology responds well to positioning within the line-of-sight range, but it cannot be positioned under non-line-of-sight conditions, and is more suitable for positioning in small-scale line-of-sight conditions; while visible light and UWB (ultra-wideband) can achieve non-line-of-sight conditions Positioning, but additional software and hardware need to be installed to assist positioning, which is costly and visible light is easily affected by other light sources; as for the indoor magnetic field, the magnetic field strength at different locations will be the same due to environmental reasons, and the positioning accuracy cannot meet the requirements.

The existing RSSI (Received Signal Strength Indication) fingerprint positioning method based on wireless signals directly calculates and compares the RSSI of the point to be positioned with the fingerprint in the fingerprint database to obtain the position coordinates of the point to be positioned or uses a neural network to realize RSSI Mapping with location coordinates. Since the fingerprint database itself is not updated in time, and the impact of device software and hardware differences on the positioning results is not considered, the real-time positioning performance is poor and the positioning accuracy of the fingerprint database is reduced.

Contents of the invention

The object of the present invention is to provide an indoor positioning method to solve the above problems. This method utilizes the structure and layout of the room to achieve regional division, so that the regional judgment is first performed during online positioning, and then the position coordinates are judged. The positioning accuracy is improved layer by layer and the positioning time is reduced; and the BP neural network is used to realize different software and hardware devices. The mapping between devices eliminates the positioning impact caused by device differences; at the same time, the fingerprint database stores multi-directional fingerprints to make positioning more directional; the introduction of a feedback mechanism enables the fingerprint database to be updated in time, and the real-time performance of the fingerprint database is enhanced.

The purpose of the present invention can adopt following technical scheme to reach:

An indoor positioning method, comprising the following steps:

Step 1. In the offline stage, according to the indoor structure and layout, divide the indoor area, establish the corresponding coordinate axes, and set the reference point in the area, and use different types of equipment to collect the RSSI of the wireless signal at the reference point to form a fingerprint vector to build the fingerprint database;

Step 2, BP neural network fitting the RSSI of the wireless signals collected by the different types of equipment;

Step 3. Build an electronic map and store the information in the database;

Step 4, in the online positioning stage, first perform regional positioning to determine the area of the point to be located; then perform position positioning in the area to determine the position coordinates of the point to be located;

Step 5. According to the target information, determine the set of reference points that need to pass through to reach the target position, and then combine the current position information to determine whether the current position is moving towards the target position;

Step 6: Update the fingerprint library in real time according to various feedback information.

Further, the specific content of constructing the fingerprint library described in step 1 includes:

1) Select the target device, and use the collected fingerprint vectors to build a fingerprint library;

2) Build a first-level fingerprint library, use a clustering algorithm for the reference points in the same area, and save the clustering results to the database as a regional mark, as a first-level fingerprint library;

3) Construct a secondary fingerprint library. For reference points in each area, select the mode of RSSI collected multiple times in different directions at the reference point as the RSSI value at the reference point. Each reference point has multiple fingerprint vectors corresponding to Multiple different directions, save it to the database as a secondary fingerprint library.

Further, the specific content of step 2 is: the number of input and output neurons of the BP neural network is equal, and its value is the number of wireless signal transmitters used for indoor positioning; wherein, the output is the selected one in step 1. The fingerprint vector collected by the target device, the input is the fingerprint vector collected by other devices.

Further, the construction of the electronic map in step 3 includes the result of area division, partial reference point information and path information between reference points.

Further, in step 1, the RSSI of the wireless signal is collected multiple times in multiple directions of each reference point;

Further, the path information between reference points indicates a set of reference points that need to pass through when arriving at a certain reference point.

Further, in step 4, the online positioning includes the following steps:

1) The collected RSSI value at the point to be located is used as input, and the mapped RSSI value is obtained through the BP neural network;

2) Regional positioning, using the nearest neighbor algorithm to calculate the mapped RSSI value and the fingerprint in the first-level fingerprint database to determine the area of the point to be located;

3) Position positioning, carry out secondary positioning in the area, use the nearest neighbor algorithm to calculate the mapped RSSI value and the fingerprint in the secondary fingerprint database, and obtain the position coordinates of the point to be located. When determining the position coordinates, the current orientation can also be obtained , and then combined with the electronic map, the location information and direction are displayed on the electronic map.

Further, the specific content of judging whether to move towards the target position in the step 5 is:

According to the reference point path information of the target position, the set of reference points to be passed is obtained, and the current position information is calculated with the reference points in the set. If the current position belongs to the set of reference points, it is moving towards the target position.

Further, in step 1, the wireless signal includes Wi-Fi, RFID and Bluetooth.

Further, in step 1, the reference point in the region is determined according to the size of the region and actual positioning accuracy requirements.

Implement the present invention, have following beneficial effect:

1. The present invention utilizes the indoor structure and layout to realize area division, so that the area is judged first during online positioning, and then the position coordinates are judged, and the positioning accuracy is progressively improved layer by layer while reducing the positioning time; and the BP neural network is used to realize different software The mapping between hardware devices eliminates the impact of positioning caused by device differences.

2. The present invention stores multi-directional fingerprints through the fingerprint library, making positioning more directional; introducing a feedback mechanism, so that the fingerprint library can be updated in time, and the real-time performance of the fingerprint library is enhanced

3. The present invention fully considers the difference of equipment software and hardware, introduces BP neural network, eliminates the influence of equipment software and hardware differences on the positioning results while positioning, and considers the impact of different directions on positioning, and collects multi-directional fingerprints for each reference point and stores them in fingerprints library; at the same time, a feedback mechanism is introduced to update the fingerprint library, and the fingerprint library has high real-time performance.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. Those skilled in the art can also obtain other drawings based on these drawings without creative work.

Fig. 1 is the schematic diagram of the off-line database construction flow chart of the indoor positioning method of the present invention;

Fig. 2 is a schematic diagram of the online positioning process of the indoor positioning method of the present invention;

3 is a schematic diagram of regional coordinates of a specific embodiment of the indoor positioning method of the present invention;

4 is a schematic diagram of an electronic map of a specific embodiment of the indoor positioning method of the present invention;

5 is a schematic diagram of an RSSI distribution histogram of a certain wireless router (WirelessPoint, WP) in an area of a specific embodiment of the indoor positioning method of the present invention;

6 is a schematic diagram of an RSSI distribution histogram of a wireless router (WP) at a reference point of a specific embodiment of the indoor positioning method of the present invention;

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

Example:

Referring to Figure 1 and Figure 2, this embodiment relates to an indoor positioning method, including the following steps:

Step 1. In the offline stage, according to the indoor structure and layout, divide the indoor area, establish the corresponding coordinate axes, and set the reference points in the area, and use different types of equipment at the reference points at multiple locations of each reference point. The RSSI of the wireless signal is collected multiple times in both directions to form a fingerprint vector to build a fingerprint database; the wireless signal includes Wi-Fi, RFID and Bluetooth. The reference point in the area is determined according to the size of the area and actual positioning accuracy requirements. The specific content of described construction fingerprint storehouse comprises:

1) Select the target device, and use the collected fingerprint vectors to build a fingerprint library;

2) Build a first-level fingerprint library, use a clustering algorithm for the reference points in the same area, and save the clustering results to the database as a regional mark, as a first-level fingerprint library;

3) Construct a secondary fingerprint library. For reference points in each area, select the mode of RSSI collected multiple times in different directions at the reference point as the RSSI value at the reference point. Each reference point has multiple fingerprint vectors corresponding to Multiple different directions, save it to the database as a secondary fingerprint library.

Step 2, BP neural network fits the RSSI of the wireless signals collected by the different types of equipment; the number of input and output neurons of the BP neural network is equal, and its value is the number of wireless signal transmitters used for indoor positioning; wherein, The output is the fingerprint vector collected by the target device selected in step 1, and the input is the fingerprint vector collected by other devices.

Step 3. Build an electronic map and store the information in a database; the electronic map includes the result of area division, some reference point information and path information between reference points. The path information between reference points indicates a set of reference points that need to pass through when reaching a certain reference point.

Step 4, in the online positioning stage, first perform regional positioning, determine the area of the point to be located; then perform position positioning in the area, and determine the position coordinates of the point to be located; the online positioning includes the following steps:

1) The collected RSSI value at the point to be located is used as input, and the mapped RSSI value is obtained through the BP neural network;

2) Regional positioning, using the nearest neighbor algorithm to calculate the mapped RSSI value and the fingerprint in the first-level fingerprint database to determine the area of the point to be located;

3) Position positioning, carry out secondary positioning in the area, use the nearest neighbor algorithm to calculate the mapped RSSI value and the fingerprint in the secondary fingerprint database, and obtain the position coordinates of the point to be located. When determining the position coordinates, the current orientation can also be obtained , and then combined with the electronic map, the location information and direction are displayed on the electronic map.

Step 5. According to the target information, determine the set of reference points that need to pass through to reach the target position, and then combine the current position information to determine whether the current position is moving towards the target position; according to the reference point path information of the target position, obtain the reference points that need to pass through Set, calculate the current position information and the reference points in the set, if the current position belongs to the set of reference points, it is moving towards the target position.

Step 6: Update the fingerprint library in real time according to various feedback information.

Specifically, the following uses Wi-Fi as a wireless signal as an example to describe in detail with the accompanying drawings:

The indoor positioning method of the present invention is divided into three stages: offline database construction, online positioning and fingerprint database update.

1. Offline database construction stage, as shown in Figure 1:

(1) According to the indoor structure and layout, there are natural shelters such as walls between different areas, and different areas are divided so that the interior of the same area is connected (without obstacles such as walls), and different areas cannot be connected (with obstacles such as walls), Assume that 5 areas can be divided, as shown in Figure 3;

(2) Coordinate the area. For the convenience of explanation, use a grid with a side length of 1 meter to divide the area, select the lower left corner as the coordinate origin, and coordinate as shown in Figure 3;

(3) Determine the reference point in the area according to the actual needs, the reference point is the grid vertex or the grid center point, as shown in Figure 3, the black point represents the reference point;

表示参考点i采集的RSSI向量，其中

表示在参考点i采集的第j个无线路由器的RSSI的值，i表示参考点总数，j为无线路由器总数；(4) At each reference point, the RSSI values of wireless routers (WP) in four orientations are collected every 2s. Assuming that there are 12 wireless routers in the room, the fingerprint vector of each reference point can be expressed as a four The tuple is AP _i = ((xi _, y _i ), D _i , A _i , RSSI _i ), where (xi _, y _i ) represents the coordinates of reference point i, and D _i represents the orientation of reference point i D _i = {0, 1, 2, 3} respectively represent the southeast and northwest, and A _i represents the area corresponding to the reference point i. A _i = {1, 2, 3, 4, 5} represents 5 different areas,

Indicates the RSSI vector collected at reference point i, where

Represent the value of the RSSI of the jth wireless router collected at reference point i, i represents the total number of reference points, and j is the total number of wireless routers;

(5) Use three different models of mobile phones for collection, and the mobile phone models are marked as A, B, and C; each reference point is collected 25 times in each direction, and each reference point is set for the RSSI value of the wireless router that cannot be collected -100dB. Such as a collection result: AP={(4,8);0;1;{-45,-39,-50,-32,-65,-100.-70,-77,-100,-100, -48,,-100,}}, indicating that the coordinates of the reference point collected in a certain time are (4,8), the orientation is east, the area number is 1, and the RSSI vector is {-45,-39,-50,-32 ,-65,-100.-70,-77,-100,-100,-48,-100,}. An example collection is as follows:

Table 1: Example of fingerprint vectors collected

model coordinate toward area RSSI A (4,8) 0 1 {-45,-39,-50,-32,-65,-100.-70,-77,-100,-100,-48,,-100,} B (4,8) 0 1 {-44,-39,-52,-34,-63,-100.-70,-77,-100,-100,-46,,-100,} ... ... ... ... ...

(6) Select the RSSI vector collected by mobile phone model A as the output of the BP neural network, and the RSSI vector collected by mobile phone models B and C as the input of the BP neural network. The number of neurons in the hidden layer is 10; the transfer function of the hidden layer neurons is selected "tansig", select "purelin" for the neuron transfer function of the output layer, select "trainingd" for the backpropagation training function, set the number of training times to 10000, and set the training target to 10 ^-7 , and use the BP neural network function that comes with MATLAB as the default Training, get the trained model and use it in the positioning phase.

(7) Using the fingerprint vector collected by the mobile phone model A as the standard, construct the fingerprint database.

(8) Construct the RSSI vector of each reference point. For the same wireless router in the same direction, we use the mode of the RSSI value in the 25 acquisitions as the RSSI value of the wireless router at the reference point, such as reference point (4,8) , facing east, collecting 25 times, the value of the wireless router is {-65,-63,-65,-65,-65,-65,-65,-65,-65,-65,-65,-65, -65,-65,-65,-65,-63,-64,-64,-66,-59,-66,-64,-64,-64}, at the reference point (4,8) , facing east, and the RSSI value of the wireless router is -65. Similarly, the RSSI vector of the wireless router corresponding to each reference point is obtained to form a secondary fingerprint library. The example of the secondary fingerprint library is as follows:

Table 2: Secondary Fingerprint Library

coordinate toward area RSSI (4,8) 0 1 {-45,-39,-50,-32,-65,-100.-70,-77,-100,-100,-48,,-100,} (4,8) 1 1 {-44,-39,-52,-34,-63,-100.-70,-77,-100,-100,-46,,-100,} ... ... ... ...

(9) For the reference points in the same area, cluster based on K-means (K-means) to obtain the first-level fingerprint library of each area. The specific process is as follows:

1) Randomly select K RSSI vectors, each RSSI vector represents a cluster center, that is, select K initial centers;

rssi^j表示参考点处测得的第j个Wi-Fi路由器的值，

表示在簇中心i处的第j个Wi-Fi路由器的RSSI的值，将它赋给与其最相似的簇中心所对应的簇；2) For each remaining RSSI vector, according to its similarity (Euclidean distance) with each cluster center,

rssi ^j represents the value of the jth Wi-Fi router measured at the reference point,

Represent the value of the RSSI of the jth Wi-Fi router at the cluster center i, and assign it to the cluster corresponding to the cluster center most similar to it;

3) Then recalculate the mean value of all RSSI vectors in each cluster as the new cluster center.

作为准则函数，即最小化每个点到最近簇中心的距离的平方和。4) Repeat 2) and 3) until the criterion function converges, that is, the cluster center does not change significantly. mean square error

As a criterion function, i.e. minimize the sum of the squares of the distances of each point to the nearest cluster center.

In the same area, the RSSI value collected by the same wireless router conforms to the Gaussian distribution, and the RSSI distribution histogram is constructed, as shown in Figure 5. According to the 2δ principle, δ is the standard deviation of the Gaussian distribution, and we can get the same wireless router in the area. The upper limit and lower limit of the RSSI value of the router, such as the upper limit is -57dB, the lower limit is -68dB, calculated according to the difference of 1dB, further determine the number of clusters in each area K = 11, the example of the first-level fingerprint library is shown in Table 3

Table 3: Example of a first-level fingerprint library

2. Build an electronic map, as shown in Figure 4;

3. Online positioning stage, as shown in Figure 2:

(1) The user holds the mobile phone assuming that the model is B, and collects the user's RSSI vector;

(2) The RSSI vector collected by the user is trained through the BP neural network to obtain a corresponding new RSSI value vector;

rssi^j表示待定位点的第j个无线路由器的值，

表示在一级指纹库中指纹i的第j个无线路由器的RSSI的值。在计算后选取距离d最小的K个点，此处取K＝1，即取查看距离d最小的点对应区域作为待测点的区域，设计算后区域判定结果取区域号1。(3) Carry out regional determination, use the KNN (K nearest neighbor) algorithm to calculate the new RSSI vector and the fingerprint in the first-level fingerprint database, and determine the user's area. In the KNN algorithm of the first-level fingerprint library, Euclidean is used as the distance judgment: that is

rssi ^j represents the value of the jth wireless router of the point to be located,

Indicates the RSSI value of the jth wireless router with fingerprint i in the first-level fingerprint database. After the calculation, select K points with the smallest distance d, where K=1 is taken, that is, the area corresponding to the point with the smallest viewing distance d is taken as the area of the point to be measured, and the area number 1 is taken as the area judgment result after the design calculation.

RSSI·RSSI_i表示待测点和参考点i的RSSI向量的点积(数量积)，

表示待测点和参考点i的RSSI向量模的乘积。每个参考点计算四次(对于四个方向)，我们取余弦相似度值最大的结果作为该点距离，说明待定位点当前朝向与其最接近，取前K个参考点计算其坐标的平均值作为待定位点的最终位置坐标，即

根据参考点处RSSI直方分布图如图6，结合高斯分布的2δ原则，δ为高斯分布的标准差，取K＝4，计算后位置坐标判定结果为(4,8)。(4) In the area whose area number is 1, use the KNN (K nearest neighbor) algorithm to perform secondary positioning to obtain the position coordinates of the point to be located. Considering the orientation in the secondary fingerprint library, the cosine similarity algorithm is used as the distance judgment:

RSSI RSSI _i represents the dot product (quantity product) of the RSSI vector of the point to be measured and the reference point i,

Indicates the product of the RSSI vector modulus of the point to be measured and the reference point i. Each reference point is calculated four times (for the four directions), and we take the result with the largest cosine similarity value as the point distance, indicating that the current orientation of the point to be located is closest to it, and take the first K reference points to calculate the average value of their coordinates As the final position coordinates of the point to be located, that is

According to the RSSI histogram distribution at the reference point as shown in Figure 6, combined with the 2δ principle of Gaussian distribution, δ is the standard deviation of Gaussian distribution, K=4, and the position coordinate judgment result after calculation is (4,8).

4. As shown in Figure 4, the current location information is 1, and the searched location information is 2, then the travel route formed by the set of reference points is the red path in Figure 2.

5. The user evaluates the positioning results. If satisfied, the RSSI value vector collected by the user is added to the corresponding position coordinates of the secondary fingerprint database, and the fingerprint database is updated. If not satisfied, the update is abandoned. RSSI collected by the user={-45,-39,-50,-32,-65,-100.-70,-77,-100,-100,-48,,-100,}, the position of the final positioning result The coordinates are (4,8), if the user is satisfied with the positioning result, the fingerprint vector {(4,8);0;1;{-45,-39,-50,-32,- 63,-100.-70,-79,-100,-100,-45,,-100,}} update to fingerprint vector {(4,8);0;1;{-45,-39,-50 ,-32,-65,-100.-70,-77,-100,-100,-48,,-100,}}.

The above disclosure is only a preferred embodiment of the present invention, which certainly cannot limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention still fall within the scope of the present invention.

Claims (7)

1. An indoor positioning method, characterized by comprising the steps of:

step 1, in an off-line stage, according to an indoor structure and layout, performing area division on an indoor environment, establishing corresponding coordinate axes, setting reference points in the area, and acquiring RSSI (received signal strength indicator) of wireless signals at the reference points by using different types of equipment to form fingerprint vectors so as to construct a fingerprint database;

step 2, the BP neural network fits the RSSI of the wireless signals collected by the different types of equipment;

step 3, constructing an electronic map, and storing the information into a database;

step 4, in the on-line positioning stage, firstly, carrying out area positioning and determining the area of a point to be positioned; then, positioning the position in the area, and determining the position coordinates of the position to be positioned;

step 5, determining a reference point set which needs to pass by to reach the target position according to the target information, and then determining whether the current position advances to the target position or not by combining the current position information;

step 6: updating the fingerprint database in real time according to various feedback information;

the specific contents of the fingerprint database construction in the step 1 comprise:

1) Selecting target equipment, and using the acquired fingerprint vector to construct a fingerprint library;

2) Constructing a primary fingerprint database, adopting a clustering algorithm for reference points in the same region, and storing a clustering result serving as a region mark in the database as the primary fingerprint database;

3) Establishing a secondary fingerprint database, selecting a mode of RSSI (received signal strength indicator) acquired for multiple times in different directions at a reference point as an RSSI value at the reference point for the reference point in each area, wherein each reference point has a plurality of fingerprint vectors corresponding to multiple different directions, and storing the fingerprint vectors into the database as the secondary fingerprint database;

in step 4, the online positioning comprises the following steps:

1) The RSSI value of the acquired point to be positioned is used as input, and the mapped RSSI value is obtained through a BP neural network;

2) Positioning the area, namely calculating the mapped RSSI value and the fingerprints in the primary fingerprint database by using a neighbor algorithm to determine the area to be positioned;

3) Performing position location, performing secondary location in the area, calculating the mapped RSSI value and fingerprints in a secondary fingerprint database by using a neighbor algorithm to obtain the position coordinates of the point to be located, determining the position coordinates and simultaneously obtaining the current orientation, and then displaying the position information and the orientation on an electronic map by combining the electronic map;

the specific content of judging whether to advance to the target position in the step 5 is as follows:

and obtaining a reference point set which needs to pass through according to the reference point path information of the target position, calculating the current position information and the reference points in the set, and if the current position belongs to the reference point set, advancing towards the target position.

2. The indoor positioning method according to claim 1, wherein the specific content of the step 2 is: the number of input and output neurons of the BP neural network is equal, and the number of the input and output neurons is the number of wireless signal transmitting ends for indoor positioning; and (3) outputting the fingerprint vector acquired by the target equipment selected in the step (1), and inputting the fingerprint vector acquired by other equipment.

3. The indoor positioning method according to claim 1, wherein the electronic map constructed in step 3 includes a result of dividing the area, information of part of the reference points, and information of paths between the reference points.

4. The indoor positioning method according to claim 1 or 2, wherein in step 1, the RSSI of the wireless signal is collected multiple times in multiple directions of each reference point.

5. The indoor positioning method according to claim 3, wherein the inter-reference-point path information indicates a set of reference points that need to be passed when reaching a certain reference point.

6. The indoor positioning method according to claim 4, wherein in step 1, the wireless signals comprise Wi-Fi, RFID and Bluetooth.

7. The indoor positioning method according to claim 1, wherein in step 1, the reference points in the area are determined according to the area size and the actual positioning accuracy requirement.

Images