Method and apparatus for classifying a probe object, storage medium, and electronic device

Abstract

The application discloses a kind of detection object classification method and device, storage medium and electronic device, the detection object classification method includes: in the case where target detection object is detected from target scene, the target radar reflection cross section (RCS) data and target track box data of target detection object are acquired;According to target RCS data and target RCS threshold corresponding to target detection object, the probability value that target detection object belongs to target type in the dimension of RCS is determined to obtain first probability, and according to track box data and track box feature data, the probability value that target detection object belongs to target type in the dimension of track box is determined to obtain second probability;According to first probability and second probability, target detection object is classified, using the above technical solution, solve the problem such as lower classification accuracy of detection object in related art.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology, and more specifically, to a method and apparatus for classifying detection objects, a storage medium, and an electronic device. Background Technology

[0002] With the development of driver assistance technologies, detection devices such as lidar deployed on driving vehicles have become crucial for data acquisition. Taking millimeter-wave radar as an example, by calculating the path time difference between the transmitted and received signals, and then through a series of signal processing methods, the distance, azimuth, and elevation angle of the detected object are obtained as measured values, which are then passed as input data to the radar data processing module. The radar data processing module correlates the measured values ​​obtained from multiple scans to obtain the target trajectory. After trajectory correlation, filtering, and other trajectory processing, the module accurately estimates the target's true trajectory data, which is then passed as the final target information to the next-level module.

[0003] Currently, in vehicular road conditions, the road environment is complex, containing both motor vehicle targets (e.g., cars, trucks) and VRU (Vulnerable Road Users) targets (e.g., pedestrians, bicycles). Pre-classification of detected objects is crucial for the correct triggering of subsequent functions. Failure to correctly classify detected objects in the road environment beforehand will reduce the accuracy of subsequent function triggering and may even lead to false alarms, causing radar misjudgments and severely impacting the radar system's object tracking performance. Therefore, accurate classification of detected objects in the radar road environment is essential. Currently, road targets can be classified by extracting the RCS (Radar Cross Section) features of detected objects during the signal processing stage. This method leverages the different RCS values ​​of different types of detected objects at different angles to extract relevant target characteristics for classification. While this approach relies on the radar's RCS value for accuracy, significant fluctuations can occur in vehicular road conditions, affecting the final classification result.

[0004] There is still no effective solution to the problem of low accuracy in classifying detected objects in related technologies. Summary of the Invention

[0005] This application provides a method and apparatus for classifying detection objects, a storage medium, and an electronic device to at least solve the problem of low accuracy in the classification of detection objects in related technologies.

[0006] According to one embodiment of this application, a method for classifying detection objects is provided, including:

[0007] When a target object is detected in the target scene, the radar cross section (RCS) data and the target track frame data of the target object are acquired. The RCS data is used to indicate the target object's ability to reflect detection waves, and the track frame data is used to indicate the track frame size of the target object.

[0008] The first probability is obtained by determining the probability value of the target detection object belonging to the target type in the RCS dimension based on the target RCS data and the target RCS threshold corresponding to the target detection object, and the second probability is obtained by determining the probability value of the target detection object belonging to the target type in the track frame dimension based on the track frame data and track frame feature data. The target RCS threshold is the RCS threshold corresponding to the location of the target detection object, and the track frame feature data is used to indicate the average features of the detection object belonging to the target type in the track frame dimension.

[0009] The target detection object is classified according to the first probability and the second probability.

[0010] Optionally, the step of determining the probability value of the target detection object belonging to the target type in the RCS dimension based on the target RCS data and the target RCS threshold corresponding to the target detection object to obtain the first probability includes:

[0011] The target RCS data is determined based on the occlusion state of the target detection object in the target scene and the reference RCS data corresponding to the target detection object, wherein the reference RCS data is the RCS detection data of the target detection object obtained from the detection device;

[0012] The target RCS threshold corresponding to the location of the target detection object is obtained from the corresponding location and RCS threshold;

[0013] The first probability is calculated based on the target RCS data and the target RCS threshold.

[0014] Optionally, determining the target RCS data based on the occlusion state of the target detection object in the target scene and the reference RCS data corresponding to the target detection object includes:

[0015] When the occlusion state indicates that the target detection object is not occluded by other detection objects in the target scene, the reference RCS data is used to determine the target RCS data;

[0016] When the occlusion state indicates that the target detection object is occluded by an associated detection object in the target scene, the reference RCS data is corrected based on the associated RCS data of the associated detection object to obtain the target RCS data. The associated detection object is a detection object that is in the same filtering channel as the target detection object, occludes the target detection object, and meets the association conditions. The association conditions include: the speed difference between the detection objects is less than the target speed threshold, and the position difference between the detection objects is greater than the target distance threshold.

[0017] Optionally, before determining the target RCS data based on the occlusion state of the target detection object in the target scene and the reference RCS data corresponding to the target detection object, the method further includes:

[0018] Acquire the position data and RCS data of all detection objects in the filtering channel where the target detection object is located;

[0019] If the location data meets the target conditions, the occlusion state is determined to be that the target detection object is occluded by a reference detection object in the target scene. The target conditions are that there is a reference detection object in the filtering channel with the same vertical position as the target detection object, and the reference detection object is located between the detection device and the target detection object, and the RCS data of the reference detection object is greater than the reference RCS data.

[0020] If the location data does not meet the target conditions, the occlusion state is determined to indicate that the target detection object is not occluded in the target scene.

[0021] Optionally, calculating the first probability based on the target RCS data and the target RCS threshold includes:

[0022] The first probability RCSscore is calculated as follows:

[0023] RCSscore=1 / (exp(-1*(RCS-RCSgate*0.9)*0.6+1)-0.1);

[0024] Wherein, RCS is the target RCS data, and RCSgate is the target RCS threshold.

[0025] Optionally, the step of determining the probability value of the target detection object belonging to the target type in the dimension of the track frame based on the track frame data and track frame feature data to obtain the second probability includes:

[0026] The bounding box length feature and bounding box length weight corresponding to the target type of the probe object are obtained, as well as the bounding box width feature and bounding box width weight corresponding to the target type of the probe object, as the bounding box feature data. The bounding box length feature is used to indicate the bounding box length attribute of the target type of the probe object, the bounding box length weight is used to indicate the degree of influence of the bounding box length attribute on the probe object belonging to the target type, the bounding box width feature is used to indicate the bounding box width attribute of the target type of the probe object, and the bounding box width weight is used to indicate the degree of influence of the bounding box width attribute on the probe object belonging to the target type.

[0027] The second probability is determined based on the track frame data and track frame feature data, wherein the track frame data includes track frame length data and track frame width data.

[0028] Optionally, determining the second probability based on the track frame data and track frame feature data includes:

[0029] The third probability wlscore is calculated as follows:

[0030] wlscore=(-1)*((T l -Mu[1])*Beta[1]+(T w -Mu[2])*Beta[2]),

[0031] The second probability wlSscore is calculated based on the third probability wlscore in the following manner:

[0032] wlSscore=1 / (exp(wlscore)+1),

[0033] Among them, T l Here, Mu[1] is the feature value of the track frame length, Beta[1] is the weight value of the track frame length, and T is the weight value of the track frame length. w Here, Mu[2] is the track frame width data, Mu[2] is the track frame width feature value, and Beta[2] is the track frame width weight value.

[0034] Optionally, obtaining the bounding box length feature and bounding box length weight corresponding to the target type of detection object, and the bounding box width feature and bounding box width weight corresponding to the target type of detection object as the bounding box feature data, includes:

[0035] Obtain the feature mean set and feature weight set corresponding to the target type of the probe object. The feature mean set records the size features of the probe object belonging to the target type in the dimension of the track frame, and the feature weight set records the feature weights corresponding to the size features of the probe object belonging to the target type in the dimension of the track frame.

[0036] The track frame length feature and the track frame width feature are obtained from the feature mean set, and the track frame length weight corresponding to the track frame length feature and the track frame width weight corresponding to the track frame width feature are obtained from the feature weight set.

[0037] Optionally, obtaining the feature mean set and feature weight set corresponding to the target type of the probe object includes:

[0038] Obtain multiple initial samples labeled with the target type to obtain an initial sample set;

[0039] For each initial sample in the initial sample set, a corresponding association feature matrix is ​​constructed to obtain the target sample set, wherein the association feature matrix includes the sample track frame length feature and the sample track frame width feature of the corresponding initial sample;

[0040] The initial support vector machine model is trained using the target sample set to obtain the target support vector machine model;

[0041] The feature mean set and the feature weight set are obtained from the target support vector machine model.

[0042] Optionally, constructing a corresponding association feature matrix for each initial sample in the initial sample set includes:

[0043] Obtain the sample frame length, sample frame width, and sample reflection cross-sectional area of ​​each initial sample, as well as the existence time of each initial sample;

[0044] The ratio between the sample track frame length and the existence time is determined as the sample track frame length feature, the ratio between the sample track frame width and the existence time is determined as the sample track frame width feature, and the ratio between the sample reflection cross-sectional area and the existence time is determined as the sample reflection cross-sectional area feature.

[0045] The sample track frame length feature, sample track frame width feature, and sample reflection cross-sectional area feature corresponding to each initial sample are used to construct the associated feature matrix corresponding to each initial sample.

[0046] Optionally, classifying the target detection object based on the first probability and the second probability includes:

[0047] The sum of the first probability and the second probability is determined as the target probability that the target detection object belongs to the target type;

[0048] If the target probability is greater than the target probability threshold, the target object is determined to belong to the target type.

[0049] If the target probability is less than or equal to the target probability threshold, it is determined that the target detection object does not belong to the target type.

[0050] According to another embodiment of the present application, a classification device for detecting objects is also provided, comprising:

[0051] The first acquisition module is used to acquire the target radar cross section (RCS) data and target track frame data of the target object when the target object is detected from the target scene. The target RCS data is used to indicate the ability of the target object to reflect the detection wave, and the target track frame data is used to indicate the track frame size of the target object.

[0052] The first determining module is used to determine a first probability by determining the probability value of the target detection object belonging to the target type in the RCS dimension based on the target RCS data and the target RCS threshold corresponding to the target detection object, and to determine a second probability by determining the probability value of the target detection object belonging to the target type in the track frame dimension based on the track frame data and track frame feature data, wherein the target RCS threshold is the RCS threshold corresponding to the location of the target detection object, and the track frame feature data is used to indicate the average features possessed by the detection object belonging to the target type in the track frame dimension;

[0053] A classification module is used to classify the target detection object according to the first probability and the second probability.

[0054] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, wherein a computer program is stored in the computer-readable storage medium, and the computer program is configured to execute the above-described classification method for the probed objects at runtime.

[0055] According to another aspect of the embodiments of this application, an electronic device is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the above-described classification method for the detected objects through the computer program.

[0056] In this embodiment, when a target object is detected in a target scene, the radar cross section (RCS) data and target track frame data of the target object are acquired. The RCS data indicates the target object's ability to reflect detection waves, and the track frame data indicates the size of the target object's track frame. A first probability is obtained by determining the probability value of the target object belonging to a target type in the RCS dimension based on the target RCS data and the target RCS threshold corresponding to the target object. A second probability is obtained by determining the probability value of the target object belonging to a target type in the track frame dimension based on the track frame data and track frame feature data. The target RCS threshold is the RCS threshold corresponding to the location of the target object, and the track frame feature data indicates the average characteristics of a target-type object in the track frame dimension. The target object is then classified according to the first and second probabilities. That is, when a target object is detected in a target scene... In this case, the radar cross section (RCS) data and target track frame data of the target object are first acquired. Since the target RCS data indicates the target object's ability to reflect detection waves, and the target track frame data indicates the size of the target object's track frame, a first probability can be obtained by determining the probability value of the target object belonging to the target type in the RCS dimension based on the target RCS data and the corresponding target RCS threshold. The target RCS threshold is the RCS threshold corresponding to the location of the target object. Furthermore, a second probability can be obtained by determining the probability value of the target object belonging to the target type in the track frame dimension based on the track frame data and track frame feature data. The track frame feature data indicates the average features possessed by the target type in the track frame dimension. Finally, the target object is classified based on both the first and second probabilities, avoiding the situation where the classification result over-reliance on RCS leads to reduced classification accuracy. This technical solution solves the problem of low classification accuracy of detected objects in related technologies, achieving the technical effect of improving the classification accuracy of detected objects. Attached Figure Description

[0057] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0058] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is a schematic diagram of the hardware environment for a method of classifying detection objects according to an embodiment of this application;

[0060] Figure 2 This is a flowchart of a method for classifying detection objects according to an embodiment of this application;

[0061] Figure 3 This is a schematic diagram of a target detection object according to an embodiment of this application;

[0062] Figure 4 This is a schematic diagram illustrating the generation of the first and second probabilities according to embodiments of this application;

[0063] Figure 5 This is a schematic diagram illustrating the generation of the first probability according to an embodiment of this application;

[0064] Figure 6 This is a schematic diagram of the filtering channel according to an embodiment of this application;

[0065] Figure 7 This is a schematic diagram illustrating the generation of the second probability according to an embodiment of this application;

[0066] Figure 8 This is a schematic diagram of a classification process for a detection object according to an embodiment of this application;

[0067] Figure 9 This is a structural block diagram of a classification device for detecting objects according to an embodiment of this application. Detailed Implementation

[0068] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0069] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0070] The methods and embodiments provided in this application can be executed on a computer terminal, device terminal, or similar computing device. Taking running on a computer terminal as an example, Figure 1 This is a schematic diagram of the hardware environment for a method for classifying detection objects according to an embodiment of this application. Figure 1 As shown, a computer terminal may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 (which may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 104 for storing data are also shown. In one exemplary embodiment, the computer terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the computer terminal described above. For example, the computer terminal may also include components that are more complex than those described above. Figure 1 The more or fewer components shown, or having the same Figure 1 Equivalent functions or ratios shown Figure 1 The functions shown have more different configurations.

[0071] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the object classification method in this embodiment of the invention. The processor 102 executes various functional applications and data processing by running the computer programs stored in the memory 104, thereby implementing the above-described method. The memory 104 may include high-speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to a computer terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0072] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by a communication provider for the computer terminal. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module used for wireless communication with the Internet.

[0073] This embodiment provides a method for classifying detection objects, applied to the aforementioned computer terminal. Figure 2 This is a flowchart of a method for classifying detection objects according to an embodiment of this application, such as... Figure 2 As shown, the process includes the following steps:

[0074] Step S202: When a target object is detected in the target scene, acquire the target radar cross section (RCS) data and target track frame data of the target object. The target RCS data is used to indicate the ability of the target object to reflect the detection wave, and the target track frame data is used to indicate the track frame size of the target object.

[0075] Step S204: Based on the target RCS data and the target RCS threshold corresponding to the target detection object, determine the probability value of the target detection object belonging to the target type in the RCS dimension to obtain a first probability, and based on the track frame data and track frame feature data, determine the probability value of the target detection object belonging to the target type in the track frame dimension to obtain a second probability. Here, the target RCS threshold is the RCS threshold corresponding to the location of the target detection object, and the track frame feature data is used to indicate the average features possessed by the detection object belonging to the target type in the track frame dimension.

[0076] Step S206: Classify the target detection object according to the first probability and the second probability.

[0077] Through the above steps, when a target object is detected in the target scene, the radar cross-section (RCS) data and target track frame data of the target object are first acquired. Since the target RCS data indicates the target object's ability to reflect detection waves, and the target track frame data indicates the size of the target object's track frame, a first probability can be obtained by determining the probability value of the target object belonging to the target type in the RCS dimension based on the target RCS data and the target RCS threshold corresponding to the target object. The target RCS threshold is the RCS threshold corresponding to the location of the target object. Furthermore, a second probability can be obtained by determining the probability value of the target object belonging to the target type in the track frame dimension based on the track frame data and track frame feature data. The track frame feature data indicates the average features possessed by the target type in the track frame dimension. Finally, the target object is classified by combining the first and second probabilities, avoiding the situation where the classification result overly relies on RCS, leading to a decrease in classification accuracy. This technical solution solves the problem of low classification accuracy of detected objects in related technologies, achieving the technical effect of improving the classification accuracy of detected objects.

[0078] In the technical solution provided in step S202 above, the target radar cross-section (RCS) data and the target track frame data may be, but are not limited to, data acquired by the detection equipment, or data obtained by processing the raw data acquired by the detection equipment. Figure 3 This is a schematic diagram of a target detection object according to an embodiment of this application, such as... Figure 3 As shown, the detection equipment deployed on the vehicle detects objects in the target scene. The detection equipment can scan the target scene by emitting detection waves. When the target object is detected, the detection data returned by the wave scan is obtained. The detection data includes the target radar cross section (RCS) data and the target track frame data. The target RCS data can indicate the target object's ability to reflect detection waves, and the target track frame data can indicate the track frame size of the target object.

[0079] In the technical solution provided in step S204 above, Figure 4 This is a schematic diagram illustrating the generation of the first and second probabilities according to embodiments of this application, as shown below. Figure 4 As shown, the first probability is determined based on the target RCS data and the target RCS threshold; the second probability is determined based on the track frame data and the track frame feature data. Therefore, the first probability may be, but is not limited to, the probability value of belonging to the target type based on the RCS dimension, and the second probability may be, but is not limited to, the probability value of belonging to the target type based on the track frame dimension.

[0080] In one exemplary embodiment, a first probability can be obtained by determining, but is not limited to, the probability value of the target detection object belonging to the target type in the RCS dimension based on the target RCS data and the target RCS threshold corresponding to the target detection object in the target scene: determining the target RCS data based on the occlusion state of the target detection object in the target scene and the reference RCS data corresponding to the target detection object, wherein the reference RCS data is the RCS detection data of the target detection object obtained from the detection device; obtaining the target RCS threshold corresponding to the location of the target detection object from the corresponding location and RCS threshold; and calculating the first probability based on the target RCS data and the target RCS threshold.

[0081] Optionally, in this embodiment, Figure 5 This is a schematic diagram illustrating the generation of the first probability according to an embodiment of this application, such as... Figure 5 As shown, by scanning the target scene with a detection device, the occlusion state of the target object in the target scene and the location of the target object in the target scene can be known. The target RCS data can be obtained based on the occlusion state and the reference RCS data obtained from the detection device. Based on the location of the target object in the target scene, the corresponding target RCS threshold can be obtained from the corresponding position and RCS threshold. Finally, the first probability can be calculated using the target RCS data and the target RCS threshold.

[0082] In one exemplary embodiment, the target RCS data can be determined, but is not limited to, in the following manner, based on the occlusion state of the target detection object in the target scene and the reference RCS data corresponding to the target detection object: when the occlusion state indicates that the target detection object is not occluded by other detection objects in the target scene, the reference RCS data is used to determine the target RCS data; when the occlusion state indicates that the target detection object is occluded by associated detection objects in the target scene, the reference RCS data is corrected based on the associated RCS data of the associated detection objects to obtain the target RCS data, wherein the associated detection objects are detection objects that are in the same filtering channel as the target detection object, occlude the target detection object, and meet the association conditions, the association conditions including: the speed difference between the detection objects is less than a target speed threshold, and the position difference between the detection objects is greater than a target distance threshold.

[0083] Optionally, in this embodiment, when the occlusion state indicates that the target detection object is occluded by an associated detection object in the target scene, the RCS detection data (reference RCS data) of the target detection object obtained by the detection device from the detection device may not be the true RCS data of the target detection object. If the reference RCS data is used directly to calculate the first probability, the accuracy of the first probability may be reduced. Therefore, the associated RCS data of the associated detection object of the target detection object is first obtained, and then the reference RCS data is corrected using the associated RCS data to obtain the target RCS data. The correction method may include, but is not limited to, assignment, such as assigning 0.8 times the associated RCS data to the reference RCS data.

[0084] Optionally, in this embodiment, the associated detection object is a detection object that is in the same filtering channel as the target detection object, obscures the target detection object, and meets the association conditions. The association conditions include: the speed difference between the detection objects is less than a target speed threshold, and the position difference between the detection objects is greater than a target distance threshold. For example, if there is a detection object A in the same filtering channel as the target detection object, and detection object A obscures the target detection object, and the speed difference between detection object A and the target detection object is less than the target speed threshold (3m / s), and the position difference between detection object A and the target detection object is greater than the target distance threshold (5m), then detection object A can be identified as an associated detection object of the target detection object.

[0085] In an exemplary embodiment, before determining the target RCS data based on the occlusion state of the target detection object in the target scene and the reference RCS data corresponding to the target detection object, the method may include, but is not limited to, obtaining the position data and RCS data of all detection objects in the filtering channel where the target detection object is located; if the position data meets the target condition, determining that the occlusion state is that the target detection object is occluded by a reference detection object in the target scene, wherein the target condition is that there is a reference detection object in the filtering channel with the same longitudinal position as the target detection object, and the reference detection object is located between the detection device and the target detection object, and the RCS data of the reference detection object is greater than the reference RCS data; if the position data does not meet the target condition, determining that the occlusion state indicates that the target detection object is not occluded in the target scene.

[0086] Optionally, in this embodiment, the filtering channel may be, but is not limited to, a method by which the detection device divides the target scene using track bars. Figure 6 This is a schematic diagram of the filtering channel according to an embodiment of this application, such as... Figure 6As shown, the target scene is divided into N regions (filter channel 1, filter channel 2, ..., filter channel n) by equidistant filter channels, and multiple detection objects existing in the target scene are assigned to their respective regions.

[0087] Optionally, in this embodiment, when the location data meets the target conditions, the occlusion state is determined to be that the target detection object is occluded by a reference detection object in the target scene. The target conditions are that there exists a reference detection object in the filtering channel with the same vertical position as the target detection object, and the reference detection object is located between the detection device and the target detection object, and the RCS data of the reference detection object is greater than the reference RCS data; for example... Figure 6 As shown, the reference detection object (detection object 2) and the target detection object (detection object 1) are in the same vertical position. The reference detection object (detection object 2) is located between the detection device and the target detection object (detection object 1), and the RCS data of the reference detection object (detection object 2) is greater than the reference RCS data. Therefore, the occlusion state is determined to be that the target detection object (detection object 1) is occluded by the reference detection object (detection object 2) in the target scene.

[0088] In one exemplary embodiment, the first probability may be calculated based on the target RCS data and the target RCS threshold in, but not limited to, the following manner:

[0089] The first probability RCSscore is calculated as follows:

[0090] RCSscore=1 / (exp(-1*(RCS-RCSgate*0.9)*0.6+1)-0.1);

[0091] Wherein, RCS is the target RCS data, and RCSgate is the target RCS threshold.

[0092] In one exemplary embodiment, a second probability can be obtained by determining, but is not limited to, the probability value of the target detection object belonging to the target type in the dimension of the track frame based on the track frame data and track frame feature data in the following manner: obtaining the track frame length feature and track frame length weight corresponding to the target type detection object, and the track frame width feature and track frame width weight corresponding to the target type detection object as the track frame feature data, wherein the track frame length feature is used to indicate the track frame length attribute of the target type detection object, the track frame length weight is used to indicate the degree of influence of the track frame length attribute on the detection object belonging to the target type, the track frame width feature is used to indicate the track frame width attribute of the target type detection object, and the track frame width weight is used to indicate the degree of influence of the track frame width attribute on the detection object belonging to the target type; determining the second probability based on the track frame data and track frame feature data, wherein the track frame data includes track frame length data and track frame width data.

[0093] Optionally, in this embodiment, Figure 7 This is a schematic diagram illustrating the generation of the second probability according to an embodiment of this application, such as... Figure 7 As shown, the second probability can be determined by track frame data and track frame feature data. The track frame data (track frame width data, track frame length data) can be obtained, but is not limited to, by detecting the target object using a detection device. The track frame feature data (track frame length feature and track frame length weight, track frame width feature and track frame width weight) can be obtained, but is not limited to, based on the target type of the target object. The track frame feature data can indicate the influence of the track frame length attribute and the track frame length attribute on the target type of the target object, as well as the influence of the track frame width attribute and the track frame width attribute on the target type of the target object.

[0094] In one exemplary embodiment, the second probability may be determined based on the track frame data and the track frame feature data, but not limited to the following:

[0095] The third probability wlscore is calculated as follows:

[0096] wlscore=(-1)*((T l -Mu[1])*Beta[1]+(T w -Mu[2])*Beta[2]),

[0097] The second probability wlSscore is calculated based on the third probability wlscore in the following manner:

[0098] wlSscore=1 / (exp(wlscore)+1),

[0099] Among them, T l Here, Mu[1] is the feature value of the track frame length, Beta[1] is the weight value of the track frame length, and T is the weight value of the track frame length. w Here, Mu[2] is the track frame width data, Mu[2] is the track frame width feature value, and Beta[2] is the track frame width weight value.

[0100] In an exemplary embodiment, the track frame length feature and track frame length weight corresponding to the target type of probe object, and the track frame width feature and track frame width weight corresponding to the target type of probe object, can be obtained as the track frame feature data by, but not limited to, the following methods: obtaining a feature mean set and a feature weight set corresponding to the target type of probe object, wherein the feature mean set records the size features of the probe object belonging to the target type in the dimension of the track frame, and the feature weight set records the feature weights corresponding to the size features of the probe object belonging to the target type in the dimension of the track frame; obtaining the track frame length feature and the track frame width feature from the feature mean set, and obtaining the track frame length weight corresponding to the track frame length feature and the track frame width weight corresponding to the track frame width feature from the feature weight set.

[0101] Optionally, in this embodiment, the feature mean set may be, but is not limited to, Mu = [Mu1, Mu2, Mu3], and the corresponding feature weight set may be, but is not limited to, Beta = [Beta1, Beta2, Beta3]. Mu1, Mu2, and Mu3 can respectively indicate the different features of the target type of the probe object in the dimension of the track box. For example, Mu1 can indicate the track box length feature, Mu2 can indicate the track box width feature, and Mu3 can indicate the track box RCS feature. The elements in the feature weight set correspond one-to-one with the elements in the feature mean set. For example, Beta1 is the weight corresponding to Mu1, Beta2 is the weight corresponding to Mu2, and so on. The track frame length feature Mu1 and the track frame width feature Mu2 can be obtained from the feature mean set Mu = [Mu1, Mu2, Mu3]. The track frame length weight Beta1 corresponding to the track frame length feature Mu1 and the track frame width weight Beta2 corresponding to the track frame width feature Mu2 can be obtained from the feature weight set Beta = [Beta1, Beta2, Beta3].

[0102] In an exemplary embodiment, the feature mean set and feature weight set corresponding to the target type of the probe object can be obtained in the following ways, but not limited to: obtaining multiple initial samples labeled with the target type to obtain an initial sample set; constructing a corresponding association feature matrix for each initial sample in the initial sample set to obtain a target sample set, wherein the association feature matrix includes the sample track frame length feature and sample track frame width feature of the corresponding initial sample; training an initial support vector machine model using the target sample set to obtain a target support vector machine model; and obtaining the feature mean set and the feature weight set from the target support vector machine model.

[0103] Optionally, in this embodiment, the target type can be, but is not limited to, any type in the target scene. It is a type set in advance according to actual needs, such as a car type. Multiple initial samples labeled with the target type are obtained, that is, multiple initial samples labeled with the car type are obtained. A corresponding association feature matrix is ​​constructed for each initial sample in the initial sample set to obtain a target sample set. An initial support vector machine model is trained using the target sample set to obtain a target support vector machine model. The target support vector machine model can accurately identify whether the input sample belongs to the car type. Then, the feature mean set Mu = [Mu1, Mu2, Mu3] and the feature weight set Beta = [Beta1, Beta2, Beta3] are obtained from the target support vector machine model.

[0104] In an exemplary embodiment, a corresponding association feature matrix may be constructed for each initial sample in the initial sample set in the following manner, but not limited to: obtaining the sample track frame length, sample track frame width, and sample reflective cross-sectional area of ​​each initial sample, as well as the existence time of each initial sample; determining the ratio between the sample track frame length and the existence time as the sample track frame length feature, the ratio between the sample track frame width and the existence time as the sample track frame width feature, and the ratio between the sample reflective cross-sectional area and the existence time as the sample reflective cross-sectional area feature; constructing the association feature matrix corresponding to each initial sample using the sample track frame length feature, the sample track frame width feature, and the sample reflective cross-sectional area feature.

[0105] Optionally, in this embodiment, the sample track frame length T of each initial sample is obtained. l Sample track frame width T w and the sample reflection cross-sectional area N s (i.e., RCS), and the existence time Age of each of the initial samples.

[0106] The sample track frame length Tl The ratio between the time of existence (Age) and the time of existence (m1) is determined as the sample track frame length feature, i.e., the sample track frame length feature m1 = T. l / Age;

[0107] The sample track frame width T w The ratio between the time of existence (Age) and the time of existence (m2) is determined as the sample track frame width feature, i.e., the sample track frame width feature m2 = T. w / Age;

[0108] The sample's reflective cross-sectional area N s The ratio between the present time Age and the sample reflectance cross-sectional area is determined as the sample reflectance cross-sectional area characteristic, that is, the sample reflectance cross-sectional area characteristic m3 = N. s / Age;

[0109] The sample track frame length feature m1, sample track frame width feature m2, and sample reflection cross-sectional area feature m3 corresponding to each initial sample are used to construct the association feature matrix M = [m1, m2, m3] corresponding to each initial sample.

[0110] In the technical solution provided in step S206 above, the target detection object is classified according to the first probability and the second probability, which avoids the situation where the classification result relies too much on RCS and thus reduces the classification accuracy.

[0111] In one exemplary embodiment, the target detection object may be classified according to the first probability and the second probability in the following manner, but not limited to: determining the sum of the first probability and the second probability as the target probability that the target detection object belongs to the target type; determining that the target detection object belongs to the target type when the target probability is greater than the target probability threshold; and determining that the target detection object does not belong to the target type when the target probability is less than or equal to the target probability threshold.

[0112] Optionally, in this embodiment, the sum of the first probability RCSscore and the second probability wlSscore is determined as the target probability that the target detection object belongs to the target type, that is, the target probability ProbValue = wlSscore + RCSscore. When the target probability ProbValue is greater than the target probability threshold G... F In the case that ProbValue > G F In the case where the target object is determined to belong to the target type; and when the target probability ProbValue is less than or equal to the target probability threshold, i.e., ProbValue ≤ G F In the case of [the target], it is determined that the target object does not belong to the target type.

[0113] To better understand the process of classifying the above-mentioned detection objects, the classification process of the above-mentioned detection objects will be described below in conjunction with optional embodiments, but this is not intended to limit the technical solutions of the embodiments of this application.

[0114] This embodiment provides a method for classifying detection objects. Figure 8 This is a schematic diagram of a classification process for a detection object according to an embodiment of this application, such as... Figure 8 As shown, the main steps include the following:

[0115] Step S801: Target point (target detection object) measurement input, the radar data processing module obtains the target information of the target detection object calculated by the signal processing module, including information such as distance, angle and velocity;

[0116] Step S802: Kalman filter estimates track information. The number of tracks (detection objects) in the target scene at time k and the track information estimate are calculated by the Kalman filter algorithm.

[0117] Step S803: Construct track bars (filter channels), traverse each track (detection object), and divide the track into different filter channels according to the vertical position of the track (detection object). Set the index ID for each filter channel;

[0118] Step S804: Determine if occluded. Before entering the support vector machine, first determine if the target object is occluded to reduce classification bias caused by inaccurate reference RCS data due to occlusion. The criteria for determining whether the target object is occluded are: there is a reference object in the filtering channel that has the same vertical position as the target object, the reference object is located between the detection device and the target object, and the RCS data of the reference object is greater than the reference RCS data;

[0119] Step S805: When the target object is in an occluded state, select the associated RCS data of the associated object to correct the reference RCS data to obtain the target RCS data. When the target object is in an unoccluded state, apply the reference RCS data and determine the probability value of the target object belonging to the target type in the RCS dimension based on the target RCS threshold of the target object to obtain the first probability.

[0120] Step S806: Enter the SVM (Support Vector Machines) target classifier training, obtain multiple initial samples labeled with the target type, and obtain an initial sample set; construct a corresponding association feature matrix for each initial sample in the initial sample set to obtain a target sample set, wherein the association feature matrix includes the sample trackbox length feature and sample trackbox width feature of the corresponding initial sample; train the initial support vector machine model using the target sample set to obtain a target support vector machine model; obtain the feature mean set and the feature weight set from the target support vector machine model, and use the feature mean set and the feature weight set to calculate the probability value of the target detection object belonging to the target type in the dimension of the trackbox to obtain a second probability;

[0121] Step S807: Obtain the car probability result through smoothing filtering. Calculate the target probability that the detected object belongs to the target type based on the first and second probabilities, and combine this with the target probability P of the detected object at time k-1. F|k-1 The probability P of the target object at time k is obtained by smoothing calculation. F|k :

[0122]

[0123] Where α and β are probability smoothing coefficients, which are empirical values ​​and can be defined, but are not limited to, as follows: α = 0.8, β = 0.2.

[0124] Step S808: Transform the target classification result through a state machine.

[0125] It should be noted that the mathematical basis of the above implementation method is the Support Vector Machine (SVM) algorithm in machine learning. Machine learning includes many algorithms, such as decision trees, random forests, perceptrons, etc., and different algorithms have different focuses and require different storage and computing resources. Although the perceptron algorithm can also output target categories when determining tracks, the storage resources required by this algorithm are greater than those of the algorithm in this paper. The embodiments of this application use a unified algorithm framework to identify target categories in the data processing stage, which does not require data fusion estimation from multiple radars. It can be easily and quickly ported to existing algorithms as an independent operating unit, and it has the advantages of clear structure, high accuracy, and low algorithm complexity.

[0126] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of this application.

[0127] Figure 9 This is a structural block diagram of a classification device for detecting objects according to an embodiment of this application; as shown below. Figure 9 As shown, it includes:

[0128] The first acquisition module 902 is used to acquire the target radar cross section (RCS) data and target track frame data of the target object when the target object is detected from the target scene. The target RCS data is used to indicate the ability of the target object to reflect the detection wave, and the target track frame data is used to indicate the track frame size of the target object.

[0129] The first determining module 904 is used to determine a first probability by determining the probability value of the target detection object belonging to the target type in the RCS dimension based on the target RCS data and the target RCS threshold corresponding to the target detection object, and to determine a second probability by determining the probability value of the target detection object belonging to the target type in the track frame dimension based on the track frame data and track frame feature data. The target RCS threshold is the RCS threshold corresponding to the location of the target detection object, and the track frame feature data is used to indicate the average features of the detection object belonging to the target type in the track frame dimension.

[0130] The classification module 906 is used to classify the target detection object according to the first probability and the second probability.

[0131] Through the above embodiments, a unified algorithm framework is used to identify target categories in the data processing stage. This eliminates the need for data fusion estimation from multiple radars and allows for convenient and quick porting to existing algorithms as an independent operating unit. Furthermore, it has advantages such as clear structure, high accuracy, and low algorithm complexity.

[0132] Through the above embodiments, when a target object is detected in a target scene, the radar cross-section (RCS) data and target track frame data of the target object are first acquired. Since the target RCS data indicates the target object's ability to reflect detection waves, and the target track frame data indicates the size of the target object's track frame, a first probability can be obtained by determining the probability value of the target object belonging to a target type in the RCS dimension based on the target RCS data and the target RCS threshold corresponding to the target object. The target RCS threshold is the RCS threshold corresponding to the location of the target object. Furthermore, a second probability can be obtained by determining the probability value of the target object belonging to a target type in the track frame dimension based on the track frame data and track frame feature data. The track frame feature data indicates the average features possessed by the target type of the detection object in the track frame dimension. Finally, the target object is classified based on both the first and second probabilities, avoiding the situation where the classification result overly relies on RCS, leading to a decrease in classification accuracy. By adopting the above technical solution, the problem of low classification accuracy of detected objects in related technologies is solved, achieving the technical effect of improving the classification accuracy of detected objects.

[0133] In an exemplary embodiment, the first determining module includes:

[0134] The first determining unit is configured to determine the target RCS data based on the occlusion state of the target detection object in the target scene and the reference RCS data corresponding to the target detection object, wherein the reference RCS data is the RCS detection data of the target detection object obtained from the detection device;

[0135] The first acquisition unit is used to acquire the target RCS threshold corresponding to the location of the target detection object from the corresponding location and RCS threshold;

[0136] The calculation unit is used to calculate the first probability based on the target RCS data and the target RCS threshold.

[0137] In an exemplary embodiment, the first determining unit is further configured to:

[0138] When the occlusion state indicates that the target detection object is not occluded by other detection objects in the target scene, the reference RCS data is used to determine the target RCS data;

[0139] When the occlusion state indicates that the target detection object is occluded by an associated detection object in the target scene, the reference RCS data is corrected based on the associated RCS data of the associated detection object to obtain the target RCS data. The associated detection object is a detection object that is in the same filtering channel as the target detection object, occludes the target detection object, and meets the association conditions. The association conditions include: the speed difference between the detection objects is less than the target speed threshold, and the position difference between the detection objects is greater than the target distance threshold.

[0140] In one exemplary embodiment, the apparatus further includes:

[0141] The second acquisition module is used to acquire the position data and RCS data of all detection objects in the filtering channel where the target detection object is located before determining the target RCS data based on the occlusion state of the target detection object in the target scene and the reference RCS data corresponding to the target detection object.

[0142] The second determining module is used to determine, when the location data meets the target conditions, that the occlusion state is that the target detection object is occluded by a reference detection object in the target scene. The target conditions are that there is a reference detection object in the filtering channel that has the same vertical position as the target detection object, and the reference detection object is located between the detection device and the target detection object, and the RCS data of the reference detection object is greater than the reference RCS data.

[0143] The third determining module is used to determine, when the location data does not meet the target conditions, that the occlusion state indicates that the target detection object is not occluded in the target scene.

[0144] In one exemplary embodiment, the computing unit is further configured to:

[0145] The first probability RCSscore is calculated as follows:

[0146] RCSscore=1 / (exp(-1*(RCS-RCSgate*0.9)*0.6+1)-0.1);

[0147] Wherein, RCS is the target RCS data, and RCSgate is the target RCS threshold.

[0148] In an exemplary embodiment, the first determining module includes:

[0149] The second acquisition unit is used to acquire the bounding box length feature and bounding box length weight corresponding to the target type of the detection object, and the bounding box width feature and bounding box width weight corresponding to the target type of the detection object as the bounding box feature data. The bounding box length feature is used to indicate the bounding box length attribute of the target type of the detection object, the bounding box length weight is used to indicate the degree of influence of the bounding box length attribute on the detection object belonging to the target type, the bounding box width feature is used to indicate the bounding box width attribute of the target type of the detection object, and the bounding box width weight is used to indicate the degree of influence of the bounding box width attribute on the detection object belonging to the target type.

[0150] The second determining unit is used to determine the second probability based on the track frame data and track frame feature data, wherein the track frame data includes track frame length data and track frame width data.

[0151] In one exemplary embodiment, the second determining unit is further configured to:

[0152] The third probability wlscore is calculated as follows:

[0153] wlscore=(-1)*((T l -Mu[1])*Beta[1]+(T w -Mu[2])*Beta[2]),

[0154] The second probability wlSscore is calculated based on the third probability wlscore in the following manner:

[0155] wlSscore=1 / (exp(wlscore)+1),

[0156] Among them, T l Here, Mu[1] is the feature value of the track frame length, Beta[1] is the weight value of the track frame length, and T is the weight value of the track frame length. w Here, Mu[2] is the track frame width data, Mu[2] is the track frame width feature value, and Beta[2] is the track frame width weight value.

[0157] In one exemplary embodiment, the second acquiring unit is further configured to:

[0158] Obtain the feature mean set and feature weight set corresponding to the target type of the probe object. The feature mean set records the size features of the probe object belonging to the target type in the dimension of the track frame, and the feature weight set records the feature weights corresponding to the size features of the probe object belonging to the target type in the dimension of the track frame.

[0159] The track frame length feature and the track frame width feature are obtained from the feature mean set, and the track frame length weight corresponding to the track frame length feature and the track frame width weight corresponding to the track frame width feature are obtained from the feature weight set.

[0160] In one exemplary embodiment, the second acquiring unit is further configured to:

[0161] Obtain multiple initial samples labeled with the target type to obtain an initial sample set;

[0162] For each initial sample in the initial sample set, a corresponding association feature matrix is ​​constructed to obtain the target sample set, wherein the association feature matrix includes the sample track frame length feature and the sample track frame width feature of the corresponding initial sample;

[0163] The initial support vector machine model is trained using the target sample set to obtain the target support vector machine model;

[0164] The feature mean set and the feature weight set are obtained from the target support vector machine model.

[0165] In one exemplary embodiment, the second acquiring unit is further configured to:

[0166] Obtain the sample frame length, sample frame width, and sample reflection cross-sectional area of ​​each initial sample, as well as the existence time of each initial sample;

[0167] The ratio between the sample track frame length and the existence time is determined as the sample track frame length feature, the ratio between the sample track frame width and the existence time is determined as the sample track frame width feature, and the ratio between the sample reflection cross-sectional area and the existence time is determined as the sample reflection cross-sectional area feature.

[0168] The sample track frame length feature, sample track frame width feature, and sample reflection cross-sectional area feature corresponding to each initial sample are used to construct the associated feature matrix corresponding to each initial sample.

[0169] In one exemplary embodiment, the classification module includes:

[0170] The third determining unit is used to determine the sum of the first probability and the second probability as the target probability that the target detection object belongs to the target type;

[0171] The fourth unit is used to determine that the target detection object belongs to the target type when the target probability is greater than the target probability threshold;

[0172] The fifth unit is used to determine that the target detection object does not belong to the target type when the target probability is less than or equal to the target probability threshold.

[0173] Embodiments of this application also provide a storage medium including a stored program, wherein the program executes any of the methods described above when it is run.

[0174] Optionally, in this embodiment, the storage medium may be configured to store program code for performing the following steps:

[0175] S1, when a target object is detected from the target scene, acquire the target radar cross section (RCS) data and target track frame data of the target object, wherein the target RCS data is used to indicate the target object's ability to reflect detection waves, and the target track frame data is used to indicate the track frame size of the target object's track frame;

[0176] S2, based on the target RCS data and the target RCS threshold corresponding to the target detection object, a first probability is obtained by determining the probability value of the target detection object belonging to the target type in the RCS dimension, and a second probability is obtained by determining the probability value of the target detection object belonging to the target type in the track frame dimension based on the track frame data and track frame feature data. Here, the target RCS threshold is the RCS threshold corresponding to the location of the target detection object, and the track frame feature data is used to indicate the average features possessed by the detection object belonging to the target type in the track frame dimension.

[0177] S3, classify the target detection object according to the first probability and the second probability.

[0178] Embodiments of this application also provide an electronic device including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0179] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0180] Optionally, in this embodiment, the processor can be configured to perform the following steps via a computer program:

[0181] S1, when a target object is detected from the target scene, acquire the target radar cross section (RCS) data and target track frame data of the target object, wherein the target RCS data is used to indicate the target object's ability to reflect detection waves, and the target track frame data is used to indicate the track frame size of the target object's track frame;

[0182] S2, based on the target RCS data and the target RCS threshold corresponding to the target detection object, a first probability is obtained by determining the probability value of the target detection object belonging to the target type in the RCS dimension, and a second probability is obtained by determining the probability value of the target detection object belonging to the target type in the track frame dimension based on the track frame data and track frame feature data. Here, the target RCS threshold is the RCS threshold corresponding to the location of the target detection object, and the track frame feature data is used to indicate the average features possessed by the detection object belonging to the target type in the track frame dimension.

[0183] S3, classify the target detection object according to the first probability and the second probability.

[0184] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0185] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and will not be repeated here.

[0186] Obviously, those skilled in the art should understand that the modules or steps of this application described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented here, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, this application is not limited to any particular combination of hardware and software.

[0187] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for classifying detection objects, characterized in that, include: When a target object is detected in the target scene, the radar cross section (RCS) data and the target track frame data of the target object are acquired. The RCS data is used to indicate the target object's ability to reflect detection waves, and the track frame data is used to indicate the track frame size of the target object. The first probability is obtained by determining the probability value of the target detection object belonging to the target type in the RCS dimension based on the target RCS data and the target RCS threshold corresponding to the target detection object, and the second probability is obtained by determining the probability value of the target detection object belonging to the target type in the track frame dimension based on the track frame data and track frame feature data. The target RCS threshold is the RCS threshold corresponding to the location of the target detection object, and the track frame feature data is used to indicate the average features of the detection object belonging to the target type in the track frame dimension. The target detection object is classified according to the first probability and the second probability; The step of determining the probability value of the target detection object belonging to the target type in the RCS dimension based on the target RCS data and the target RCS threshold corresponding to the target detection object to obtain the first probability includes: determining the target RCS data based on the occlusion state of the target detection object in the target scene and the reference RCS data corresponding to the target detection object, wherein the reference RCS data is the RCS detection data of the target detection object obtained from the detection device; obtaining the target RCS threshold corresponding to the location of the target detection object from the corresponding location and RCS threshold; and calculating the first probability based on the target RCS data and the target RCS threshold. The step of determining the probability value of the target detection object belonging to the target type in the dimension of the track frame based on the track frame data and track frame feature data to obtain the second probability includes: obtaining the track frame length feature and track frame length weight corresponding to the target type detection object, and the track frame width feature and track frame width weight corresponding to the target type detection object as the track frame feature data, wherein the track frame length feature is used to indicate the track frame length attribute of the target type detection object, the track frame length weight is used to indicate the degree of influence of the track frame length attribute on the detection object belonging to the target type, the track frame width feature is used to indicate the track frame width attribute of the target type detection object, and the track frame width weight is used to indicate the degree of influence of the track frame width attribute on the detection object belonging to the target type; and determining the second probability based on the track frame data and track frame feature data, wherein the track frame data includes track frame length data and track frame width data.

2. The method according to claim 1, characterized in that, The step of determining the target RCS data based on the occlusion state of the target detection object in the target scene and the reference RCS data corresponding to the target detection object includes: When the occlusion state indicates that the target detection object is not occluded by other detection objects in the target scene, the reference RCS data is used to determine the target RCS data; When the occlusion state indicates that the target detection object is occluded by an associated detection object in the target scene, the reference RCS data is corrected based on the associated RCS data of the associated detection object to obtain the target RCS data. The associated detection object is a detection object that is in the same filtering channel as the target detection object, occludes the target detection object, and meets the association conditions. The association conditions include: the speed difference between the detection objects is less than the target speed threshold, and the position difference between the detection objects is greater than the target distance threshold.

3. The method according to claim 1, characterized in that, Before determining the target RCS data based on the occlusion state of the target detection object in the target scene and the reference RCS data corresponding to the target detection object, the method further includes: Obtain the position data and RCS data of all detection objects in the filtering channel where the target detection object is located; If the location data meets the target conditions, the occlusion state is determined to be that the target detection object is occluded by a reference detection object in the target scene. The target conditions are that there is a reference detection object in the filtering channel with the same vertical position as the target detection object, and the reference detection object is located between the detection device and the target detection object, and the RCS data of the reference detection object is greater than the reference RCS data. If the location data does not meet the target conditions, the occlusion state is determined to indicate that the target detection object is not occluded in the target scene.

4. The method according to claim 1, characterized in that, The step of calculating the first probability based on the target RCS data and the target RCS threshold includes: The first probability RCSscore is calculated as follows: ； in, For the target RCS data, The target RCS threshold is defined as follows.

5. The method according to claim 1, characterized in that, Determining the second probability based on the track frame data and track frame feature data includes: The third probability is calculated in the following way. : ， According to the third probability The second probability is calculated in the following manner. : ， Among them, among them, The length of the track frame is the data. The feature value of the track frame length. The weight value for the track frame length. The width of the track frame. The feature value of the track frame width. The weight value for the width of the track frame.

6. The method according to claim 1, characterized in that, The step of obtaining the bounding box length feature and bounding box length weight corresponding to the target type of detection object, and the bounding box width feature and bounding box width weight corresponding to the target type of detection object as the bounding box feature data includes: Obtain the feature mean set and feature weight set corresponding to the target type of the probe object. The feature mean set records the size features of the probe object belonging to the target type in the dimension of the track frame, and the feature weight set records the feature weights corresponding to the size features of the probe object belonging to the target type in the dimension of the track frame. The track frame length feature and the track frame width feature are obtained from the feature mean set, and the track frame length weight corresponding to the track frame length feature and the track frame width weight corresponding to the track frame width feature are obtained from the feature weight set.

7. The method according to claim 6, characterized in that, The acquisition of the feature mean set and feature weight set corresponding to the target type of the probe object includes: Obtain multiple initial samples labeled with the target type to obtain an initial sample set; For each initial sample in the initial sample set, a corresponding association feature matrix is ​​constructed to obtain the target sample set, wherein the association feature matrix includes the sample track frame length feature and the sample track frame width feature of the corresponding initial sample; The initial support vector machine model is trained using the target sample set to obtain the target support vector machine model; The feature mean set and the feature weight set are obtained from the target support vector machine model.

8. The method according to claim 7, characterized in that, The step of constructing a corresponding association feature matrix for each initial sample in the initial sample set includes: Obtain the sample frame length, sample frame width, and sample reflection cross-sectional area of ​​each initial sample, as well as the existence time of each initial sample; The ratio between the sample track frame length and the existence time is determined as the sample track frame length feature, the ratio between the sample track frame width and the existence time is determined as the sample track frame width feature, and the ratio between the sample reflection cross-sectional area and the existence time is determined as the sample reflection cross-sectional area feature. The sample track frame length feature, sample track frame width feature, and sample reflection cross-sectional area feature corresponding to each initial sample are used to construct the associated feature matrix corresponding to each initial sample.

9. The method according to claim 1, characterized in that, The classification of the target detection object based on the first probability and the second probability includes: The sum of the first probability and the second probability is determined as the target probability that the target detection object belongs to the target type; If the target probability is greater than the target probability threshold, the target object is determined to belong to the target type. If the target probability is less than or equal to the target probability threshold, it is determined that the target detection object does not belong to the target type.

10. A classification device for detecting objects, characterized in that, include: The first acquisition module is used to acquire the target radar cross section (RCS) data and target track frame data of the target object when the target object is detected from the target scene. The target RCS data is used to indicate the ability of the target object to reflect the detection wave, and the target track frame data is used to indicate the track frame size of the target object. The first determining module is used to determine a first probability by determining the probability value of the target detection object belonging to the target type in the RCS dimension based on the target RCS data and the target RCS threshold corresponding to the target detection object, and to determine a second probability by determining the probability value of the target detection object belonging to the target type in the track frame dimension based on the track frame data and track frame feature data, wherein the target RCS threshold is the RCS threshold corresponding to the location of the target detection object, and the track frame feature data is used to indicate the average features possessed by the detection object belonging to the target type in the track frame dimension; A classification module is used to classify the target detection object according to the first probability and the second probability; The first determining module includes: a first determining unit, configured to determine the target RCS data based on the occlusion state of the target detection object in the target scene and the reference RCS data corresponding to the target detection object, wherein the reference RCS data is the RCS detection data of the target detection object obtained from the detection device; a first acquiring unit, configured to acquire the target RCS threshold corresponding to the location of the target detection object from a corresponding location and RCS threshold; and a calculation unit, configured to calculate the first probability based on the target RCS data and the target RCS threshold. The first determining module includes: a second acquiring unit, configured to acquire the bounding box length feature and bounding box length weight corresponding to the target type of the probe object, and the bounding box width feature and bounding box width weight corresponding to the target type of the probe object as the bounding box feature data, wherein the bounding box length feature is used to indicate the bounding box length attribute of the target type of the probe object, the bounding box length weight is used to indicate the degree of influence of the bounding box length attribute on the probe object belonging to the target type, the bounding box width feature is used to indicate the bounding box width attribute of the target type of the probe object, and the bounding box width weight is used to indicate the degree of influence of the bounding box width attribute on the probe object belonging to the target type; and a second determining unit, configured to determine the second probability based on the bounding box data and the bounding box feature data, wherein the bounding box data includes bounding box length data and bounding box width data.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein the program, when executed, performs the method of any one of claims 1 to 9.

12. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to execute the method of any one of claims 1 to 9 through the computer program.

Images