CN110146865B - Target identification method and device for radar image - Google Patents
Target identification method and device for radar image Download PDFInfo
- Publication number
- CN110146865B CN110146865B CN201910471034.1A CN201910471034A CN110146865B CN 110146865 B CN110146865 B CN 110146865B CN 201910471034 A CN201910471034 A CN 201910471034A CN 110146865 B CN110146865 B CN 110146865B
- Authority
- CN
- China
- Prior art keywords
- signal
- target
- point cloud
- probability
- noise ratio
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000000875 corresponding effect Effects 0.000 claims description 123
- 238000009826 distribution Methods 0.000 claims description 75
- 238000000034 method Methods 0.000 claims description 47
- 230000033001 locomotion Effects 0.000 claims description 46
- 238000012937 correction Methods 0.000 claims description 21
- 238000012545 processing Methods 0.000 claims description 12
- 238000012216 screening Methods 0.000 claims description 10
- 230000002596 correlated effect Effects 0.000 claims description 5
- 230000009467 reduction Effects 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 8
- 230000008569 process Effects 0.000 description 8
- 238000010801 machine learning Methods 0.000 description 4
- 230000003247 decreasing effect Effects 0.000 description 2
- 238000012217 deletion Methods 0.000 description 2
- 230000037430 deletion Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000002592 echocardiography Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000000149 penetrating effect Effects 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 230000008685 targeting Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
- G01S7/414—Discriminating targets with respect to background clutter
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
- G01S7/415—Identification of targets based on measurements of movement associated with the target
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/48—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S17/00
- G01S7/4802—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S17/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Radar Systems Or Details Thereof (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the specification provides a target recognition method and device for radar images. In the process of target identification aiming at signal-to-noise ratio information carried by signal points of a radar image, on one hand, the signal points in the radar image are initially filtered through a lower first preset threshold value to obtain an initial point cloud image, so that data processing capacity can be primarily screened and reduced, and on the other hand, the probability distribution point cloud of each target on the current radar image is predicted through historical motion state information of each target by utilizing a historical target identification result. Further, the signal-to-noise ratio information of each signal point in the initial point cloud image is corrected through probability distribution of each target, and the target point cloud image is obtained through screening out the signal points with signal-to-noise ratios lower than the standard threshold in the corrected initial point cloud image, so that the targets in the current radar image are identified from the target point cloud image. This embodiment can improve the effectiveness of target recognition.
Description
Technical Field
One or more embodiments of the present disclosure relate to the field of computer technology, and more particularly, to a method and apparatus for identifying objects in radar images by a computer.
Background
The object identifier is used for identifying an object from one or more images or videos through a computer, and can also identify the state of the object, such as direction, speed and the like. Target recognition is often achieved by means of a camera to collect visual images, or by means of a point cloud image collected by reflection of radar (laser, microwave, etc.), or the like. In the conventional technology, under the condition that target identification is carried out through a point cloud image acquired by a radar, a threshold value is often set, the signal-to-noise ratio of a signal received by the radar is filtered, a point cloud image which is a target is identified for a signal point cluster with the signal-to-noise ratio higher than the threshold value of a reflected signal, and the target is identified, positioned and tracked. For signal points where the signal-to-noise ratio of the reflected signal is below the threshold, it is considered noise and filtered out.
However, in some situations, if the object density in the detection area is too high, the object to be detected may be blocked by other objects, and even if the penetrability of the radar can penetrate the object to reflect through the object, the signal-to-noise ratio of the reflected signal of the real object will be greatly reduced. In the conventional art, if the threshold is lowered, a high noise signal point may be identified as a target; if the threshold value is increased, signal points corresponding to the real target may be filtered out, so that the target cannot be positioned and tracked.
Disclosure of Invention
One or more embodiments of the present specification describe a method and apparatus for target recognition of radar images to address one or more of the problems mentioned above.
According to a first aspect, there is provided a method of object recognition, the method comprising: acquiring a current radar image of a preset area, wherein the radar image comprises a plurality of signal points, and each signal point carries signal-to-noise ratio information of a corresponding radar reflection signal; screening out signal points with signal to noise ratio lower than a first preset threshold value from the plurality of signal points, so as to obtain an initial point cloud image corresponding to the current radar image, wherein the first preset threshold value is lower than a preset standard threshold value; predicting each probability distribution point cloud of each target on a current radar image by utilizing corresponding historical motion state information aiming at each target in a historical target identification result in the preset area, wherein each probability distribution point cloud describes the occurrence probability of each target at a corresponding position through a plurality of probability points; correcting the signal-to-noise ratio information of corresponding signal points in the initial point cloud image according to each probability distribution point cloud; and screening out signal points with signal to noise ratio lower than the standard threshold value from the corrected initial point cloud image to obtain a target point cloud image aiming at the preset area so as to identify a target in the current radar image from the target point cloud image.
In one embodiment, the signal-to-noise ratio information of the radar-reflected signal corresponding to each signal point is represented by a gray value of the corresponding signal point, wherein the gray value of each signal point is proportional to the corresponding radar-reflected signal.
In one embodiment, each object in the historical object recognition results within the predetermined region includes a first object; the predicting each probability distribution point cloud of each target on the current radar image by using the corresponding historical motion state information comprises: acquiring each piece of motion state information in a previous preset historical target identification result of the first target; and inputting the motion state information into a pre-trained prediction model, and determining the probability distribution of the first target at the current radar image acquisition moment according to the output result of the prediction model.
In one embodiment, the historical motion state information includes at least historical location information, and further includes one or more of: direction, angle, speed.
In one embodiment, probabilities represented by respective probability points in the probability distribution point cloud are gaussian.
In one embodiment, the historical target recognition result in the predetermined area comprises a plurality of targets; the correcting the signal-to-noise ratio information of the corresponding signal points in the initial point cloud image according to each probability distribution point cloud comprises the following steps: superposing the probability distribution point clouds corresponding to the targets to generate a superposition point cloud; and determining a first probability point which is consistent with the first signal point in the superimposed point cloud for the first signal point in the initial point cloud image, and correcting the signal-to-noise ratio information corresponding to the first signal point according to a preset correction rule based on the first probability corresponding to the first probability point.
In one embodiment, the correction rule includes: reducing the signal-to-noise ratio corresponding to the first signal point according to a preset signal-to-noise ratio reduction mode under the condition that the first probability is smaller than a preset probability threshold; and under the condition that the first probability is larger than the preset probability threshold, increasing the signal to noise ratio corresponding to the first signal point according to a preset signal to noise ratio increasing mode, wherein in the preset signal to noise ratio increasing mode, a signal to noise ratio increasing value or a corrected value is positively correlated with the first probability.
In one embodiment, the method further comprises: clustering the signal points in the target point cloud image according to the corrected signal-to-noise ratio information and the coordinate position information; and respectively identifying each point cluster obtained by clustering as each target in the current radar image.
In one embodiment, the method further comprises: and storing a target recognition result corresponding to the current radar image so as to be used for target recognition of the subsequent radar image.
According to a second aspect, there is provided an apparatus for object recognition, the apparatus comprising:
the radar image acquisition unit is configured to acquire a current radar image of a preset area, wherein the radar image comprises a plurality of signal points, and each signal point carries signal-to-noise ratio information of a corresponding radar reflection signal;
The first processing unit is configured to screen out signal points with signal to noise ratios lower than a first preset threshold value from the plurality of reflected signal points, so that an initial point cloud image corresponding to the current radar image is obtained, wherein the first preset threshold value is lower than a preset standard threshold value;
the second processing unit is configured to predict each probability distribution point cloud of each target on the current radar image by utilizing corresponding historical motion state information aiming at each target in the historical target recognition result in the preset area, wherein each probability distribution point cloud describes the occurrence probability of each target at a corresponding position through a plurality of probability points;
the correcting unit is configured to correct the signal-to-noise ratio information of corresponding signal points in the initial point cloud image according to each probability distribution point cloud;
and the identification unit is configured to screen out signal points with signal to noise ratios lower than the standard threshold value from the corrected initial point cloud image to obtain a target point cloud image aiming at the preset area so as to identify a target in the current radar image from the target point cloud image.
According to a third aspect, there is provided a computer readable storage medium having stored thereon a computer program which, when executed in a computer, causes the computer to perform the method of the first aspect.
According to a fourth aspect, there is provided a computing device comprising a memory and a processor, characterised in that the memory has executable code stored therein, the processor implementing the method of the first aspect when executing the executable code.
According to the target recognition method and device for the radar image, provided by the embodiment of the specification, in the process of target recognition of the radar image through signal-to-noise ratio information carried by signal points, on one hand, the signal points in the radar image are initially filtered through a lower first preset threshold value to obtain an initial point cloud image, so that data processing amount can be primarily screened and reduced, and on the other hand, the historical target recognition result is utilized, and the probability distribution point cloud of each target on the current radar image is predicted through the historical motion state information of each target. Further, the signal-to-noise ratio information of each signal point in the initial point cloud image is corrected through probability distribution of each target, and the target point cloud image is obtained through screening out the signal points with signal-to-noise ratios lower than the standard threshold in the corrected initial point cloud image, so that the targets in the current radar image are identified from the target point cloud image. Therefore, the situation that the signal-to-noise ratio of the blocked target corresponding to the signal point on the radar image is low and erroneous deletion and erroneous judgment are caused can be prevented, and the effectiveness of target identification is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 shows a schematic diagram of an implementation scenario of an embodiment of the present description;
FIG. 2 illustrates a flow of a method for target recognition of radar images, according to one embodiment;
FIG. 3 is a schematic diagram of a probability distribution point cloud corresponding to a target when the target is a human body in a specific example;
FIG. 4 shows a schematic diagram of probability distribution of individual probability points in a probability distribution point cloud in one specific example;
fig. 5 shows a schematic block diagram of an apparatus for target recognition of radar images according to one embodiment.
Detailed Description
The following describes the scheme provided in the present specification with reference to the drawings.
Fig. 1 is a schematic diagram of an implementation scenario of the embodiment of the present specification. In this implementation scenario, a predetermined area, a radar signal acquisition platform, a computing platform are included. The radar signal acquisition platform acquires radar reflection signals in a preset area, and the computing platform processes the signals to identify targets in the preset area, so that radar images at corresponding moments are acquired. The computing platform may be various hardware devices with certain computing capabilities or a software system running on a hardware device, etc., such as a desktop computer, a server, etc. It will be appreciated that the computing platform may also be a cluster of devices, e.g., devices as described above. The computing platform may be disposed in a predetermined area, or may be disposed at a remote server, which is not limited herein.
It is noted that the radar signal collection platform may be any device capable of collecting radar signals. The device and the computing platform can be of an integral structure or can be arranged separately. The radar signal collection platform may include, for example, a radar transmitter and a receiver, where the transmitter transmits a lidar or microwave (e.g., millimeter wave) radar to a predetermined area, and the receiver collects signals reflected back from objects within the predetermined area. The reflected signal may be referred to herein as a radar reflected signal. At some point, the reflected signals of various objects within the predetermined area form a radar image.
The objects in the predetermined area may be distributed or may be overlapped. When objects are arranged in an overlapping manner, the target identified by the reflected signal of the radar may not be sufficiently comprehensive. When the target is identified, the computing platform firstly carries out target identification on the current radar image, a coarser initial point cloud image containing noise is obtained by setting a lower threshold value, then the signal correction is carried out on the initial point cloud image at the current moment through the motion state information of the corresponding target in the historical point cloud image, the target identification is carried out on the corrected point cloud image, and the information such as the target of the preset area, the motion state of the target and the like is determined.
In practice, radar images may have a certain acquisition time interval, which is typically related to the emission properties of the radar (e.g. the interval of one frame). The radar emits electromagnetic waves of a certain frequency at predetermined emission intervals (e.g., 40 ms), and receives reflected signals passing through objects within a predetermined area, forming a radar image. The transmission interval may also be the acquisition interval of radar images. For the radar images acquired at each moment, the computing platform can execute the target identification method provided by the embodiment of the specification to identify the targets and the motion state information of the targets. Specifically, if the acquisition time corresponding to the current radar image is recorded by T, and the radar image at the T time is filtered by coarse-granularity signals to obtain an initial point cloud image at the T time, the computing platform can perform signal correction on the initial point cloud image at the T time through the history point cloud image between T-T and the T time. Where T is a predetermined time interval and may comprise a predetermined number of frames (e.g., 5 frames) of the point cloud image. In some implementations, the historical point cloud image for signal correction may also be determined directly from the predetermined number of frames. Thus, at least from the point cloud image of the predetermined frame (the predetermined frame may be preceded by a predetermined frame number, such as the 6 th frame), each frame of the point cloud image may be subjected to correction of the point cloud image of the predetermined frame number, and the target identification is performed through an adaptive iterative process. The point cloud image of a predetermined frame number (e.g., 5 frames) before the first predetermined frame may be processed in a predetermined manner, which is not described herein. The target identification mode can effectively avoid the situation that part of targets are missed under the condition that the targets are shielded by objects and the like.
One problem worth explaining is that the object recognition method of the embodiment of the present specification uses the historical motion state information, and therefore, the object recognition method for radar images of the present specification is particularly suitable for the recognition process of moving objects. For a temporarily stationary object, the motion state of the object may also be considered as a stationary state.
The process for target recognition of radar images is described in detail below.
FIG. 2 illustrates a method flow diagram for object recognition, according to one embodiment. The execution main body of the method shown in fig. 2 is, for example, the computing platform in fig. 1. As shown in fig. 2, the method of object recognition includes the steps of: step 202, acquiring a current radar image of a preset area, wherein the radar image comprises a plurality of signal points, and each signal point carries signal-to-noise ratio information of a corresponding radar reflection signal; step 204, screening out signal points with signal to noise ratio lower than a first preset threshold from the plurality of signal points, so as to obtain an initial point cloud image corresponding to the current radar image, wherein the first preset threshold is lower than a preset standard threshold; step 206, predicting each probability distribution point cloud of each target on the current radar image by utilizing corresponding historical motion state information according to each target of the historical recognition result in the preset area, wherein each probability distribution point cloud describes the occurrence probability of each target at a corresponding position through a plurality of probability points; step 208, correcting the signal-to-noise ratio information of the corresponding signal points in the initial point cloud image according to each probability distribution point cloud; step 210, for the corrected initial point cloud image, screening out signal points with signal-to-noise ratios lower than the standard threshold to obtain a target point cloud image for a predetermined area, so as to identify a target in the current radar image from the target point cloud image.
First, in step 202, a radar image of a predetermined area is acquired. The predetermined area may be an area within the radar monitoring range, such as a room, all areas of radar signal radiation, etc.
It will be appreciated that the radar may illuminate the target by emitting electromagnetic waves and receive echoes thereof, thereby obtaining information about the distance of the target from the point of emission of the electromagnetic waves, the rate of change of distance (e.g. radial velocity, etc.), the bearing, altitude, etc. The echo information here may also be referred to as a reflected signal. When multiple reflected signals are clustered together, a radar image is depicted. A reflected signal may correspond to a signal point on the radar image.
The signal-to-noise ratio (SNR) is a parameter describing the proportional relationship between the active component and the noise component in a signal, for example, the ratio of the power of the active component to the power of the noise component. In the radar reflected signal, the effective component may be a signal component reflected from the target object, and the noise component may be an air particle reflected signal, a reflected signal generated by a penetrating object where the radar reaches the target, and an interference signal component such as other signals generated by the device. The signal to noise ratio may take on a value between 0 and 1. The closer the signal-to-noise ratio is to 0, the greater the component of the noise signal, the closer the signal-to-noise ratio is to 1, and the greater the specific gravity of the effective signal. Thus, signal points with large noise signal components can be filtered out by a suitable signal-to-noise threshold.
In practice, the signal-to-noise ratio information described above may be carried by signal points on the radar image. For example, the signal-to-noise ratio of the corresponding radar-reflected signal may be described by the gray value of the signal point. The gray value may be proportional to the signal-to-noise ratio. For example, the gray value range of the signal points on the radar image is 0-255, which may correspond to signal to noise ratios of 0-1, respectively. In one implementation, the gray value determination method is, for example: the 256 x signal to noise ratio is rounded off and then reduced by 1.
When the flow shown in fig. 2 is used for real-time processing of the radar image, the current radar image may be the radar image at the current time. While the flow shown in fig. 2 is used for batch processing of historical radar images, the current radar image may also be a radar image that is currently acquired in time order.
Then, through step 204, signal points with signal-to-noise ratios lower than a first preset threshold are screened out from the plurality of signal points, so as to obtain an initial point cloud image corresponding to the current radar image. It will be appreciated that in determining a point cloud image of a target from a radar image, screening may be performed according to a pre-set signal-to-noise threshold, referred to herein as a standard threshold. That is, after the standard threshold is determined, a signal point having a signal-to-noise ratio higher than the standard threshold is regarded as a point forming the target point cloud, and a signal point having a signal-to-noise ratio lower than the standard threshold is regarded as a noise point. The purpose of this step 204 is to perform a preliminary screening on the reflected signal points in the radar image, filter out the signal points with a higher probability of being noise points, and retain the signal points that are likely to be misjudged as noise points due to the fact that the signal-to-noise ratio is low due to the shielding between objects, so a lower signal-to-noise ratio threshold can be selected. That is, the first preset threshold is lower than the preset standard threshold.
The first preset threshold value can be determined manually and empirically, or can be determined by a machine learning method. Taking a machine learning method as an example, the signal to noise ratios of the signal points screened out by the history can be clustered according to the method, the signal to noise ratios of the signal points screened out by the history can be clustered by taking M and L as centers respectively, so that the signal points screened out by the history are clustered into 2 categories, wherein the clustering category taking M as the center corresponds to the signal point which is misjudged as the noise point because of low signal to noise ratio, and the clustering category taking L as the center corresponds to the noise point. The initial value of M can be N×3/4, and the initial value of L can be N×1/4. If the signal point truly misjudged as the noise point is lower than a certain proportion (such as 90%) in the clustering class taking the current value of M as the center, M is moved from the initial value to N by a preset step length (such as 0.01N), and clustering is carried out again. And (3) until the signal points truly misjudged as noise points in the clustering class with M as the center reach the certain proportion, and taking the last clustering result. If the signal point truly misjudged as the noise point is higher than the certain proportion in the clustering class taking the current value of M as the center, M is moved from the initial value to N/2 by a preset step length (such as 0.01N), and clustering is carried out again. And (3) until the signal point which is truly misjudged as the noise point is lower than the certain proportion in the clustering class taking M as the center, taking the previous clustering result. And then, regarding the obtained clustering result, taking the minimum signal-to-noise ratio corresponding to each signal point in the clustering category taking M as a center as a first preset threshold.
On the other hand, for each target in the historical recognition result in the predetermined area, each probability distribution point cloud of each target on the current radar image is predicted by using the corresponding historical motion state information, through step 206. The respective targets may be the targets in the target recognition results of the radar images of the preceding predetermined frame number in succession, or the targets in the target recognition results of the radar images in the preceding predetermined period, and are not limited herein. One or more targets may be included in the target recognition result corresponding to each radar image.
It will be appreciated that for each target it may occur in one target recognition result or in a plurality of target recognition results. The object motion has continuity, and for a continuously moving object, the motion state at the next moment can be estimated according to the historical motion state.
The reflection characteristics for radar are also different for different objects. Thus, object recognition can be performed according to different information carried by signal points in the radar image. For example, the clustering of the signal points according to the signal-to-noise ratio and the coordinates corresponding to the signal points may be performed. The coordinates of the object reflection point corresponding to the signal point may be polar coordinates with the radar emission point as an origin, the distance between the object reflection point and the origin may be described by the radial coordinates, and the angle of the object reflection point relative to the origin may be described by the angular coordinates. Coordinate information may also be carried in signal points of the radar image. In this way, object reflection points corresponding to the clusters obtained by each cluster in the clustering result are continuous in space and can correspond to one target. The object here may be, for example, a person, a vehicle, or the like.
It will be appreciated that the reflection characteristics of the radar may be different at different locations of the same target, and in some implementations, only specific portions of the target may be identified in order to avoid interference between the targets. For example, when the target is a person, the radar may be set at a higher position (such as a roof or the like), and the different persons may be distinguished by collecting and identifying signal points reflected from the positions of the head or the like of the person. Therefore, under the conditions of partial mutual shielding among human bodies and the like, the influence of deviation of clustering (such as that different human bodies with relatively close distances gather into a cluster) on the target recognition effect can be avoided.
In the radar image, since the radar signal has periodicity and continuity, the motion state of each signal point, such as the current speed, the motion direction, and the like, can also be determined in the radar image. For one target, there is also an overall state of motion.
In one embodiment, predictive models of historical and current motion states may be pre-trained. The training samples of the prediction model may be multiple groups of target images, each group of target images corresponds to one target, the group of target images includes R target images which are continuously arranged according to a preset time interval, the motion states of the targets in the first R-1 target images in each group are sequentially input into the selected model, and the model parameters are adjusted according to the motion states of the targets in the R-th target image, so that the prediction model is trained. In the target recognition flow shown in fig. 2, assuming that any target in the historical target recognition results in the predetermined area is the first target, each piece of motion state information of the first target in the predetermined (previous 5) historical target recognition results can be acquired, the motion state information is input into the pre-trained prediction model, and the output result of the prediction model is that the first target corresponds to the estimated position of the current radar image. These estimated positions may be described in terms of probability distributions. In other words, the first target may be described by a certain probability corresponding to each position appearing on the current radar image. The probability that a target appears at a corresponding location may also be referred to as a probability point. Multiple probability points may form a probability distribution point cloud image.
As an example, please refer to fig. 3, which is a schematic diagram of probability distribution point clouds of a target on a current radar image. As shown in fig. 3, the probability distribution of the estimated position of a human body at a certain time in the scene targeting the human body is shown. Wherein, the color depth represents probability, the darkest part is the estimated position and the motion state of the human body at the moment, and the corresponding distribution probability value is larger. The probability of the distribution of the lightest colored portions is close to 0. The probability distribution may be used to describe the probability of the distribution of the target over a predetermined radius, centered on the overall estimated position of the target.
In another embodiment, the motion state of the target can be predicted according to the continuous motion rule of the target and the parameters such as direction, speed and the like. For example, in a specific example, the target is advanced at a constant speed according to the historical motion state, and assuming that the coordinates of a certain point in the target at the time T are (r, θ), the estimated coordinates of the point at the time t+t are predicted to be (r+v) r t，θ+V θ t). Wherein r and θ are the radial coordinate and the angular coordinate of the point at the moment T, and V r 、V θ The movement speed of the point along the radial direction and the circumferential direction at the moment T respectively, and T can be the time interval of radar image acquisition. The estimated position and the motion state of the target can be determined through the estimated coordinates of each point corresponding to the target. When determining the probability distribution of the target at the predetermined position, it can be considered that each point coordinate calculated in the above manner is a point where the probability of occurrence of the target is large.
On the probability distribution image of the first object, the probability of occurrence of the object may be described by the gray value of each point, or the probability of occurrence of the object may be described by the probability value. When the probability distribution is gaussian, the points with higher probability of occurrence of the target are concentrated near the estimated position. As shown in fig. 4, probability points with a larger probability of target distribution are concentrated on probability points corresponding to the region 401, and probability points at other positions (such as probability points corresponding to the region 402) have a smaller probability of target occurrence. At this time, the sum of probabilities of occurrence of the objects described at the respective probability points in the probability distribution image of the first object may also be 1. By the method, probability distribution point clouds of all targets on the current radar image in the historical target recognition result can be determined.
Further, in step 208, the probability distribution point cloud of each target determined in step 206 may be used to correct the signal-to-noise ratio information of the corresponding signal point in the initial point cloud image obtained in step 204. It can be appreciated that the probability distribution point cloud of each target delineates the probability of each location point of the target in the current radar image, and the target identification process is determined based on the signal-to-noise ratio of the corresponding signal point in the current radar image, so that the signal-to-noise ratio of the corresponding signal point in the initial point cloud image can be modified based on each probability distribution.
And under the condition that a target is contained in the historical target identification result in the preset area, distributing the point cloud according to the probability corresponding to the target, and correcting the signal-to-noise ratio information of each signal point in the initial point cloud image. And under the condition that a plurality of targets are contained in the historical target identification result in the preset area, each probability distribution point cloud can be overlapped, and the signal-to-noise ratio of the corresponding reflection signal point in the initial point cloud image is corrected by the overlapped probability distribution point cloud. In some implementations, normalization may also be performed on the superimposed probability distribution point cloud to ensure that the sum of probabilities corresponding to the respective points in the superimposed probability distribution image is 1.
Taking the case that a plurality of targets are contained in a historical target identification result in a predetermined area as an example, assuming that any signal point in an initial point cloud image is a first signal point, a first probability point consistent with the first signal point in the superimposed point cloud can be determined first, and signal-to-noise ratio information corresponding to the first signal point is corrected according to a predetermined correction rule based on a first probability corresponding to the first probability point.
According to one embodiment, the correction rule may be a signal-to-noise correction value that converts probabilities corresponding to each probability point in the probability distribution point cloud into corresponding reflected signal points in the initial point cloud image. The sum of the signal-to-noise value of the corresponding signal point in the initial point cloud image and the correction value can be used as the corrected signal-to-noise value. For example, the probability value is multiplied by a coefficient, so that the sum of the maximum signal-to-noise ratio in the initial point cloud image and the maximum probability value in the probability distribution image does not exceed 1, and the product of the probability value of each point and the coefficient can be used as the signal-to-noise ratio correction value of the corresponding reflection signal point.
In another embodiment, the correction rule may include: under the condition that the first probability is smaller than a preset probability threshold value, reducing the signal-to-noise ratio corresponding to the first signal point according to a preset signal-to-noise ratio reduction mode; and under the condition that the first probability is larger than a preset probability threshold value, increasing the signal to noise ratio corresponding to the first signal point according to a preset signal to noise ratio increasing mode. Wherein, in the preset signal-to-noise ratio increasing mode, the signal-to-noise ratio increasing value or the corrected value is positively correlated with the first probability, and in the preset signal-to-noise ratio decreasing mode, the signal-to-noise ratio decreasing value or the corrected value is negatively correlated with the first probability.
Taking the probability distribution point cloud as an example, the probability corresponding to each probability point is in gaussian distribution, and the probability points with larger occurrence probability of the target are concentrated. As shown in fig. 4, the probability value distribution at this time has a significant fall between the region 401 and the region 402. If the probability values (preset probability threshold) corresponding to the boundary between the region 401 and the region 402 are determined, the signal-to-noise ratio of the corresponding signal point is increased for the probability points corresponding to each probability value in the region 401, and the signal-to-noise ratio of the corresponding signal point is reduced for the probability points corresponding to each probability value in the region 402.
In this way, the signal-to-noise ratio value of the signal point corresponding to the probability point with the target occurrence probability larger than the preset probability threshold can be further increased, and the signal-to-noise ratio value of the signal point corresponding to the probability point with the target occurrence probability smaller than the preset probability threshold can be further reduced, so that the gap is enlarged, and the high noise point is reduced.
It should be noted that, since the probability distribution point cloud is estimated according to the target historical motion track, the points in the probability distribution point cloud and the signal points in the initial point cloud image do not necessarily correspond strictly. Further, in determining the correspondence between probability points and reflected signal points in the initial point cloud image, it is not necessarily an exact point-to-point correspondence.
In one embodiment, the points in the probability distribution point cloud may be first divided into regions, where the points corresponding to the same probability are divided into the same region. Taking a certain divided area as an example, the area corresponds to a corresponding area of the initial point cloud image, and the signal points in the corresponding area are taken as the signal points consistent with the probability points in the area. The signal to noise ratio information of the signal points can be corrected according to the probability corresponding to the probability points in the area.
In another embodiment, individual signal points in the initial point cloud image may be mapped onto the probability distribution point cloud image. Taking a certain signal point as an example, after mapping to a first position of a probability distribution point cloud, determining a probability point closest to the first position on a probability distribution point cloud image, determining the probability point as a point consistent with the signal point, and correcting signal-to-noise ratio information of the signal point according to the probability corresponding to the probability point.
In more embodiments, when determining the correspondence between the probability points in the probability distribution point cloud and the signal points in the initial point cloud image, the determination may be performed in other manners, which will not be described herein.
Next, in step 210, for the corrected initial point cloud image, the signal points with signal-to-noise ratios lower than the standard threshold are screened out to obtain a target point cloud image for the predetermined area, so as to identify the target in the current radar image from the target point cloud image. The standard threshold may be determined empirically or based on a machine learning model. Taking machine learning as an example, signal points in a plurality of radar images containing targets can be taken as samples, wherein the signal points corresponding to the targets are positive samples, other signal points (noise points) are negative samples, and clustering of the positive and negative samples containing two clustering categories according to signal-to-noise ratio values is performed by means of clustering and the like. The two cluster categories correspond to a positive sample category and a negative sample category, respectively. Under the condition that the clustering result distinguishes the true positive and negative samples according to a certain accuracy (such as 95%), the minimum signal-to-noise ratio value in the clustering category corresponding to the positive sample can be used as a standard threshold.
In some embodiments, since one or more targets may be included in the initial point cloud image, the target identification may be further performed on the point cloud image after the reflected signal points with signal-to-noise ratios below the standard threshold are screened out. The object recognition process may be performed by a clustering method. Specifically, signal points in the target point cloud image can be clustered according to the corrected signal-to-noise ratio information and the coordinate position information, and each point cluster in the clustering result is identified as a target in the current radar image.
In a further embodiment, the target recognition result corresponding to the current radar image may also be stored for target recognition of the subsequent radar image.
Reviewing the above process, in the process of carrying out target recognition on the signal-to-noise ratio information carried by the radar image through the signal points, on one hand, the signal points in the radar image are initially filtered through a lower first preset threshold value to obtain an initial point cloud image, so that the data processing amount can be primarily screened and reduced, and on the other hand, the historical target recognition result is utilized, and the probability distribution point cloud of each target on the current radar image is predicted through the historical motion state information of each target. Further, the signal-to-noise ratio information of each signal point in the initial point cloud image is corrected through probability distribution of each target, and the target point cloud image is obtained through screening out the signal points with signal-to-noise ratios lower than the standard threshold in the corrected initial point cloud image, so that the targets in the current radar image are identified from the target point cloud image. Therefore, the situation that the signal-to-noise ratio of the blocked target corresponding to the signal point on the radar image is low and erroneous deletion and erroneous judgment are caused can be prevented, and the effectiveness of target identification is improved.
According to an embodiment of another aspect, there is also provided an object recognition apparatus for radar images. Fig. 5 shows a schematic block diagram of an object recognition device for radar images according to one embodiment. As shown in fig. 5, the object recognition apparatus 500 for radar image includes:
an obtaining unit 51, configured to obtain a current radar image of a predetermined area, where the radar image includes a plurality of signal points, and each signal point carries signal-to-noise ratio information of a corresponding radar reflection signal;
the first processing unit 52 is configured to screen out signal points with a signal-to-noise ratio lower than a first preset threshold from the plurality of reflected signal points, so as to obtain an initial point cloud image corresponding to the current radar image, wherein the first preset threshold is lower than a preset standard threshold;
a second processing unit 53 configured to predict, for each target in the historical target recognition result in the predetermined area, each probability distribution point cloud of each target on the current radar image by using the corresponding historical motion state information, each probability distribution point cloud describing probability of occurrence of each target at a corresponding position by a plurality of probability points, respectively;
a correction unit 54 configured to correct the signal-to-noise ratio information of the corresponding signal points in the initial point cloud image according to each probability distribution point cloud;
And an identifying unit 55 configured to screen out signal points with signal-to-noise ratios lower than the standard threshold for the corrected initial point cloud image to obtain a target point cloud image for a predetermined area, so as to identify a target in the current radar image from the target point cloud image.
In one embodiment, the signal-to-noise ratio information of the radar-reflected signal corresponding to each signal point is represented by a gray value of the corresponding signal point, wherein the gray value of each signal point is proportional to the corresponding radar-reflected signal.
Assuming that each target in the history recognition result within the predetermined area includes a first target, the second processing unit 53 may be further configured to:
acquiring each piece of motion state information in a previous preset historical target identification result of a first target;
and inputting each piece of motion state information into a pre-trained prediction model, and determining probability distribution of the first target at the current radar image acquisition moment according to the output result of the prediction model.
According to one embodiment, the historical motion state information includes at least historical position information, and further includes one or more of the following: direction, angle, speed.
In one embodiment, probabilities represented by respective probability points in the probability distribution point cloud are gaussian.
In the case where a plurality of targets are included in the history target recognition result within the predetermined area, the correction unit 54 may be further configured to:
superposing the probability distribution point clouds corresponding to the targets to generate a superposition point cloud;
and determining a first probability point which is consistent with the first signal point in the superimposed point cloud for the first signal point in the initial point cloud image, and correcting the signal-to-noise ratio information corresponding to the first signal point according to a preset correction rule based on the first probability corresponding to the first probability point.
In a further embodiment, the correction rule may include:
under the condition that the first probability is smaller than a preset probability threshold value, reducing the signal-to-noise ratio corresponding to the first signal point according to a preset signal-to-noise ratio reduction mode;
and under the condition that the first probability is larger than a preset probability threshold, increasing the signal to noise ratio corresponding to the first signal point according to a preset signal to noise ratio increasing mode, wherein in the preset signal to noise ratio increasing mode, the signal to noise ratio increasing value or the corrected value is positively related to the first probability.
According to one embodiment, the identification unit 55 may be further configured to:
clustering the signal points in the target point cloud image according to the corrected signal-to-noise ratio information and the coordinate position information;
And respectively identifying each point cluster obtained by each cluster as each target in the current radar image.
According to one possible design, the apparatus 500 may further comprise an updating unit configured to:
and storing a target recognition result corresponding to the current radar image so as to be used for target recognition of the subsequent radar image.
It should be noted that, the apparatus 500 shown in fig. 5 is an apparatus embodiment corresponding to the method embodiment shown in fig. 2, and the corresponding description in the method embodiment shown in fig. 2 is also applicable to the apparatus 500, which is not repeated herein.
According to an embodiment of another aspect, there is also provided a computer-readable storage medium having stored thereon a computer program which, when executed in a computer, causes the computer to perform the method described in connection with fig. 2.
According to an embodiment of yet another aspect, there is also provided a computing device including a memory having executable code stored therein and a processor that, when executing the executable code, implements the method described in connection with fig. 2.
Those skilled in the art will appreciate that in one or more of the examples described above, the functions described in the present invention may be implemented in hardware, software, firmware, or any combination thereof. When implemented in software, these functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium.
The foregoing embodiments have been provided for the purpose of illustrating the general principles of the present invention in further detail, and are not to be construed as limiting the scope of the invention, but are merely intended to cover any modifications, equivalents, improvements, etc. based on the teachings of the invention.
Claims (20)
1. A target recognition method for a radar image, the method comprising:
acquiring a current radar image of a preset area, wherein the radar image comprises a plurality of signal points, and each signal point carries signal-to-noise ratio information of a corresponding radar reflection signal;
screening out signal points with signal to noise ratio lower than a first preset threshold value from the plurality of signal points, so as to obtain an initial point cloud image corresponding to the current radar image, wherein the first preset threshold value is lower than a preset standard threshold value;
predicting each probability distribution point cloud of each target on a current radar image by utilizing corresponding historical motion state information aiming at each target in a historical recognition result in the preset area, wherein each probability distribution point cloud describes the occurrence probability of each target at a corresponding position through a plurality of probability points;
Correcting the signal-to-noise ratio information of corresponding signal points in the initial point cloud image according to each probability distribution point cloud;
and screening out signal points with signal to noise ratio lower than the standard threshold value from the corrected initial point cloud image to obtain a target point cloud image aiming at the preset area so as to identify a target in the current radar image from the target point cloud image.
2. The method of claim 1, wherein the signal-to-noise ratio information of the radar-reflected signal corresponding to each signal point is represented by a gray value of the corresponding signal point, wherein the gray value of each signal point is proportional to the corresponding radar-reflected signal.
3. The method of claim 1, wherein each object within the predetermined area comprises a first object;
the predicting each probability distribution point cloud of each target on the current radar image by using the corresponding historical motion state information comprises:
acquiring each piece of motion state information in a previous preset historical target identification result of the first target;
and inputting the motion state information into a pre-trained prediction model, and determining the probability distribution of the first target at the current radar image acquisition moment according to the output result of the prediction model.
4. A method according to claim 1 or 3, wherein the historical motion state information comprises at least historical location information, further comprising one or more of: direction, angle, speed.
5. The method of claim 1, wherein probabilities represented by individual probability points in the probability distribution point cloud are gaussian.
6. The method of claim 1, wherein the historical target recognition result in the predetermined area comprises a plurality of targets;
the correcting the signal-to-noise ratio information of the corresponding signal points in the initial point cloud image according to each probability distribution point cloud comprises the following steps:
superposing the probability distribution point clouds corresponding to the targets to generate a superposition point cloud;
and determining a first probability point which is consistent with the first signal point in the superimposed point cloud for the first signal point in the initial point cloud image, and correcting the signal-to-noise ratio information corresponding to the first signal point according to a preset correction rule based on the first probability corresponding to the first probability point.
7. The method of claim 6, wherein the correction rule comprises:
reducing the signal-to-noise ratio corresponding to the first signal point according to a preset signal-to-noise ratio reduction mode under the condition that the first probability is smaller than a preset probability threshold;
And under the condition that the first probability is larger than the preset probability threshold, increasing the signal to noise ratio corresponding to the first signal point according to a preset signal to noise ratio increasing mode, wherein in the preset signal to noise ratio increasing mode, a signal to noise ratio increasing value or a corrected value is positively correlated with the first probability.
8. The method of claim 1, wherein the method further comprises:
clustering the signal points in the target point cloud image according to the corrected signal-to-noise ratio information and the coordinate position information;
and respectively identifying each point cluster obtained by clustering as each target in the current radar image.
9. The method of claim 1, wherein the method further comprises:
and storing a target recognition result corresponding to the current radar image so as to be used for target recognition of the subsequent radar image.
10. An object recognition device for radar images, the device comprising:
the radar image acquisition unit is configured to acquire a current radar image of a preset area, wherein the radar image comprises a plurality of signal points, and each signal point carries signal-to-noise ratio information of a corresponding radar reflection signal;
the first processing unit is configured to screen out signal points with signal to noise ratios lower than a first preset threshold value from the plurality of signal points so as to obtain an initial point cloud image corresponding to the current radar image, wherein the first preset threshold value is lower than a preset standard threshold value;
The second processing unit is configured to predict each probability distribution point cloud of each target on the current radar image by utilizing corresponding historical motion state information aiming at each target in the historical target recognition result in the preset area, wherein each probability distribution point cloud describes the occurrence probability of each target at a corresponding position through a plurality of probability points;
the correcting unit is configured to correct the signal-to-noise ratio information of corresponding signal points in the initial point cloud image according to each probability distribution point cloud;
and the identification unit is configured to screen out signal points with signal to noise ratios lower than the standard threshold value from the corrected initial point cloud image to obtain a target point cloud image aiming at the preset area so as to identify a target in the current radar image from the target point cloud image.
11. The apparatus of claim 10, wherein signal-to-noise ratio information of the radar-reflected signal corresponding to each signal point is represented by a gray value of the corresponding signal point, wherein the gray value of each signal point is proportional to the corresponding radar-reflected signal.
12. The apparatus of claim 10, wherein each object within the predetermined area comprises a first object;
The second processing unit is further configured to:
acquiring each piece of motion state information in a previous preset historical target identification result of the first target;
and inputting the motion state information into a pre-trained prediction model, and determining the probability distribution of the first target at the current radar image acquisition moment according to the output result of the prediction model.
13. The apparatus of claim 10 or 12, wherein the historical motion state information includes at least historical location information, further comprising one or more of: direction, angle, speed.
14. The apparatus of claim 10, wherein probabilities represented by individual probability points in the probability distribution point cloud are gaussian.
15. The apparatus of claim 10, wherein the historical target recognition result in the predetermined area comprises a plurality of targets;
the correction unit is further configured to:
superposing the probability distribution point clouds corresponding to the targets to generate a superposition point cloud;
and determining a first probability point which is consistent with the first signal point in the superimposed point cloud for the first signal point in the initial point cloud image, and correcting the signal-to-noise ratio information corresponding to the first signal point according to a preset correction rule based on the first probability corresponding to the first probability point.
16. The apparatus of claim 15, wherein the correction rule comprises:
reducing the signal-to-noise ratio corresponding to the first signal point according to a preset signal-to-noise ratio reduction mode under the condition that the first probability is smaller than a preset probability threshold;
and under the condition that the first probability is larger than the preset probability threshold, increasing the signal to noise ratio corresponding to the first signal point according to a preset signal to noise ratio increasing mode, wherein in the preset signal to noise ratio increasing mode, a signal to noise ratio increasing value or a corrected value is positively correlated with the first probability.
17. The apparatus of claim 10, wherein the identification unit is further configured to:
clustering the signal points in the target point cloud image according to the corrected signal-to-noise ratio information and the coordinate position information;
and respectively identifying each point cluster obtained by clustering as each target in the current radar image.
18. The apparatus of claim 10, wherein the apparatus further comprises an updating unit:
and storing a target recognition result corresponding to the current radar image so as to be used for target recognition of the subsequent radar image.
19. A computer readable storage medium having stored thereon a computer program which, when executed in a computer, causes the computer to perform the method of any of claims 1-9.
20. A computing device comprising a memory and a processor, wherein the memory has executable code stored therein, which when executed by the processor, implements the method of any of claims 1-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910471034.1A CN110146865B (en) | 2019-05-31 | 2019-05-31 | Target identification method and device for radar image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910471034.1A CN110146865B (en) | 2019-05-31 | 2019-05-31 | Target identification method and device for radar image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110146865A CN110146865A (en) | 2019-08-20 |
| CN110146865B true CN110146865B (en) | 2023-07-14 |
Family
ID=67590178
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910471034.1A Active CN110146865B (en) | 2019-05-31 | 2019-05-31 | Target identification method and device for radar image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110146865B (en) |
Families Citing this family (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111091148B (en) * | 2019-12-10 | 2021-07-27 | 成都纳雷科技有限公司 | Vehicle-mounted radar target tracking method and device based on target prior information |
| CN111158014B (en) * | 2019-12-30 | 2023-06-30 | 华通科技有限公司 | Multi-radar comprehensive bird detection system |
| CN111781608B (en) * | 2020-07-03 | 2023-04-25 | 浙江光珀智能科技有限公司 | Moving target detection method and system based on FMCW laser radar |
| CN112285665B (en) * | 2020-10-14 | 2023-12-26 | 广州极飞科技股份有限公司 | Signal detection method and device, millimeter wave radar module, equipment and medium |
| CN115436893B (en) * | 2021-06-01 | 2025-07-01 | 富士通株式会社 | Key point correction device and method based on wireless radar signal |
| CN113793307B (en) * | 2021-08-23 | 2024-11-08 | 上海派影医疗科技有限公司 | An automatic annotation method and system suitable for multiple types of pathological images |
| CN113657331A (en) * | 2021-08-23 | 2021-11-16 | 深圳科卫机器人科技有限公司 | Warning line infrared induction identification method and device, computer equipment and storage medium |
| CN113920778A (en) * | 2021-12-15 | 2022-01-11 | 深圳佑驾创新科技有限公司 | Image acquisition method and device |
| WO2023108544A1 (en) * | 2021-12-15 | 2023-06-22 | 深圳航天科技创新研究院 | Single-antenna ultra-wideband radar system for imaging application |
| CN116363857A (en) * | 2021-12-27 | 2023-06-30 | 深圳成谷软件有限公司 | Traffic flow detection method, device, electronic equipment and storage medium |
| CN115410408B (en) * | 2022-08-31 | 2024-05-24 | 南京慧尔视智能科技有限公司 | Parking space state change detection method, device, equipment and medium |
| CN116626630B (en) * | 2023-07-25 | 2023-09-29 | 北京赛目科技股份有限公司 | Object classification method and device, electronic equipment and storage medium |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017143217A1 (en) * | 2016-02-18 | 2017-08-24 | Aeye, Inc. | Adaptive ladar receiver |
Family Cites Families (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100565244C (en) * | 2007-10-31 | 2009-12-02 | 北京航空航天大学 | Multi-mode multi-target precision tracking device and method |
| JP2017156219A (en) * | 2016-03-02 | 2017-09-07 | 沖電気工業株式会社 | Tracking device, tracking method, and program |
| CN106054169B (en) * | 2016-05-18 | 2018-09-21 | 西安电子科技大学 | Multistation Radar Signal Fusion detection method based on tracking information |
| KR101752858B1 (en) * | 2016-12-09 | 2017-07-19 | 메타빌드주식회사 | Radar-based high-precision unexpected incident detecting system |
| CN108152831B (en) * | 2017-12-06 | 2020-02-07 | 中国农业大学 | Laser radar obstacle identification method and system |
| CN108230379B (en) * | 2017-12-29 | 2020-12-04 | 百度在线网络技术（北京）有限公司 | Method and device for fusing point cloud data |
| CN108509918B (en) * | 2018-04-03 | 2021-01-08 | 中国人民解放军国防科技大学 | Target detection and tracking method fusing laser point cloud and image |
| CN109100741B (en) * | 2018-06-11 | 2020-11-20 | 长安大学 | A target detection method based on 3D lidar and image data |
| CN108469609B (en) * | 2018-06-11 | 2022-02-18 | 成都纳雷科技有限公司 | Detection information filtering method for radar target tracking |
| CN111666832B (en) * | 2018-07-27 | 2023-10-31 | 创新先进技术有限公司 | Detection method and device, computing equipment and storage medium |
-
2019
- 2019-05-31 CN CN201910471034.1A patent/CN110146865B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017143217A1 (en) * | 2016-02-18 | 2017-08-24 | Aeye, Inc. | Adaptive ladar receiver |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110146865A (en) | 2019-08-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110146865B (en) | Target identification method and device for radar image | |
| US10007836B2 (en) | Bird detection device, bird detection system, bird detection method, and program extracting a difference between the corrected images | |
| JP7111175B2 (en) | Object recognition system, recognition device, object recognition method, and object recognition program | |
| CN111624567B (en) | Constant false alarm detection method and device | |
| CN112101069B (en) | Method and device for determining driving area information | |
| CN111323756B (en) | Marine radar target detection method and device based on deep learning | |
| CN110400294B (en) | Infrared target detection system and detection method | |
| CN114139373B (en) | Multi-sensor automatic collaborative management method for unmanned aerial vehicle reverse vehicle | |
| CN107274410B (en) | Self-adaptive artificial target constant false alarm rate detection method | |
| CN114355960B (en) | Unmanned aerial vehicle defense intelligent decision-making method and system, server and medium | |
| EP3893020A1 (en) | System for processing radar data representing intensity values of received power of reflected radar wave signals | |
| CN111323757B (en) | Target detection method and device for marine radar | |
| CN109493366A (en) | Based on background modeling method, system and the medium for visiting bird radar image | |
| CN111784750A (en) | Method, device, device and storage medium for tracking moving objects in video images | |
| CN112698295B (en) | Knowledge-assisted radar detection and tracking integrated method and system | |
| CN110827257B (en) | Visual navigation positioning method for embedded airborne infrared image | |
| CN112836571A (en) | Ship target detection and recognition method, system and terminal in remote sensing SAR images | |
| EP4357806A1 (en) | A method and device for filtering radar reflection points | |
| US20220383146A1 (en) | Method and Device for Training a Machine Learning Algorithm | |
| CN117111624A (en) | Anti-unmanned aerial vehicle method and system based on electromagnetic anti-control technology | |
| US12002371B2 (en) | Neuromorphic cameras for aircraft | |
| WO2023193923A1 (en) | Maritime traffic management | |
| Sánchez-Oro et al. | Urban traffic surveillance in smart cities using radar images | |
| Pereira et al. | Infrared and Visible Camera Integration for Detection and Tracking of Small UAVs: Systematic Evaluation. | |
| CN119224750B (en) | A SAR Image Target Detection and Tracking Method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20200924 Address after: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant after: Innovative advanced technology Co.,Ltd. Address before: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant before: Advanced innovation technology Co.,Ltd. Effective date of registration: 20200924 Address after: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant after: Advanced innovation technology Co.,Ltd. Address before: A four-storey 847 mailbox in Grand Cayman Capital Building, British Cayman Islands Applicant before: Alibaba Group Holding Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |