Radar track classification methods, devices, and storage media

Abstract

This application provides a method, apparatus, and storage medium for classifying radar tracks. The method includes: acquiring radar echo data; and obtaining radar track classification results based on the radar echo data using a radar track classification model. The radar track classification method, apparatus, and storage medium provided in this application generate radar echo images and structured features such as time, position, and velocity from radar echo signals. These features are then processed by a target detection and tracking network to generate track sequence feature vectors. Radar track classification is performed based on these feature vectors. This approach integrates multiple feature information and extracts sequence features, fully leveraging the characteristics of echo data and improving the accuracy of radar track classification and detection.

Description

Technical Field

[0001] This application relates to the field of radar technology, and in particular to a method, apparatus and storage medium for classifying radar tracks. Background Technology

[0002] Radar uses radio methods to detect targets and determine their spatial location. Based on vast amounts of radar echo data, radar can perform multiple tasks. The primary task of modern radar is target detection and identification. The main challenge for radar today is, under given system conditions, to fully mine echo information, improve radar target detection accuracy through software algorithms, and complete related tasks such as track classification and track correlation, thus propelling radar systems towards intelligence and information technology.

[0003] Existing radar track classification research mainly utilizes spectral information, polarization features, and high-resolution range profiles from radar echoes to identify targets in a single frame. However, radar echo images carry limited information features, and traditional track classification methods extract relatively little information from radar echoes, resulting in low information utilization and consequently, low accuracy in track classification and detection. Summary of the Invention

[0004] This application provides a radar track classification method, apparatus, and storage medium to solve the technical problem of low accuracy in radar track classification and detection in related technologies.

[0005] In a first aspect, embodiments of this application provide a method for classifying radar tracks, including:

[0006] Acquire radar echo data;

[0007] Based on the radar echo data, a radar track classification model is used to obtain the radar track classification result. The radar track classification model is obtained by training on the track sequence information obtained through echo image information, echo time information, echo position information and target velocity information, based on the track sequence feature vector and the track category label corresponding to the track sequence feature vector.

[0008] In some embodiments, the training steps of the radar track classification model include:

[0009] Based on the radar echo data, the feature vector of the track point is determined;

[0010] Extract feature vectors from multiple track points to determine the track sequence feature vector;

[0011] The radar track classification model is trained based on the track sequence feature vector and the track category label corresponding to the track sequence feature vector.

[0012] In some embodiments, determining the feature vector of the track point based on the radar echo data includes:

[0013] Based on the radar echo data, the echo image information, echo time information, echo position information, and target velocity information of a single frame of radar echo are obtained.

[0014] The echo image information, echo time information, echo position information, and target velocity information of the single-frame radar echo are fused to form the track point feature vector.

[0015] In some embodiments, acquiring echo image information, echo time information, echo position information, and target velocity information of a single-frame radar echo based on the radar echo data includes:

[0016] The radar echo data is analyzed to obtain echo image information, echo time information, and echo location information;

[0017] Target detection is performed on the echo image information, and target velocity information is obtained using a target tracking algorithm.

[0018] In some embodiments, the step of performing target detection on the echo image information and obtaining target velocity information using a target tracking algorithm includes:

[0019] Target detection is performed on the echo image information based on the target detection algorithm to obtain the radar echo target;

[0020] Based on the target tracking algorithm, the trajectory sequence image is obtained from the radar echo target, and the target speed information is calculated.

[0021] In some embodiments, fusing the echo image information, echo time information, echo position information, and target velocity information of the single-frame radar echo to form the track point feature vector includes:

[0022] Image features corresponding to the echo image information of a single-frame radar echo are obtained based on a three-layer fully connected network model, and the structured information corresponding to the echo time information, echo position information and target velocity information of a single-frame radar echo is obtained based on a one-layer fully connected network model.

[0023] By fusing the image features and the features of the structured information, a track point feature vector is obtained.

[0024] In some embodiments, the step of obtaining the image features corresponding to the echo image information of a single-frame radar echo based on a three-layer fully connected network model includes:

[0025] The echo image information of a single-frame radar echo is preprocessed and then input into the three-layer fully connected network model to obtain the image features corresponding to the echo image information of the single-frame radar echo output by the three-layer fully connected network model; the activation function of the three-layer fully connected network model is a non-linear activation function.

[0026] In some embodiments, extracting feature vectors from multiple waypoints to determine the feature vector of the waypoint sequence includes:

[0027] By concatenating the feature vectors of multiple track points, multiple candidate track sequence feature vectors are obtained;

[0028] The feature vector of the candidate track sequence containing the most waypoints is determined as the track sequence feature vector.

[0029] Secondly, embodiments of this application also provide a radar track classification device, comprising:

[0030] The first acquisition module is used to acquire radar echo data;

[0031] The first processing module is used to obtain radar track classification results based on the radar echo data using a radar track classification model. The radar track classification model is obtained by training on track sequence feature vectors and track category labels corresponding to track sequence feature vectors, based on track sequence feature vectors and track sequence feature vectors.

[0032] In some embodiments, the training steps of the radar track classification model include:

[0033] Based on the radar echo data, the feature vector of the track point is determined;

[0034] Extract feature vectors from multiple track points to determine the track sequence feature vector;

[0035] The radar track classification model is trained based on the track sequence feature vector and the track category label corresponding to the track sequence feature vector.

[0036] In some embodiments, determining the feature vector of the track point based on the radar echo data includes:

[0037] Based on the radar echo data, the echo image information, echo time information, echo position information, and target velocity information of a single frame of radar echo are obtained.

[0038] The echo image information, echo time information, echo position information, and target velocity information of the single-frame radar echo are fused to form the track point feature vector.

[0039] In some embodiments, acquiring echo image information, echo time information, echo position information, and target velocity information of a single-frame radar echo based on the radar echo data includes:

[0040] The radar echo data is analyzed to obtain echo image information, echo time information, and echo location information;

[0041] Target detection is performed on the echo image information, and target velocity information is obtained using a target tracking algorithm.

[0042] In some embodiments, the step of performing target detection on the echo image information and obtaining target velocity information using a target tracking algorithm includes:

[0043] Target detection is performed on the echo image information based on the target detection algorithm to obtain the radar echo target;

[0044] Based on the target tracking algorithm, the trajectory sequence image is obtained from the radar echo target, and the target speed information is calculated.

[0045] In some embodiments, fusing the echo image information, echo time information, echo position information, and target velocity information of the single-frame radar echo to form the track point feature vector includes:

[0046] Image features corresponding to the echo image information of a single-frame radar echo are obtained based on a three-layer fully connected network model, and the structured information corresponding to the echo time information, echo position information and target velocity information of a single-frame radar echo is obtained based on a one-layer fully connected network model.

[0047] By fusing the image features and the features of the structured information, a track point feature vector is obtained.

[0048] In some embodiments, the step of obtaining the image features corresponding to the echo image information of a single-frame radar echo based on a three-layer fully connected network model includes:

[0049] The echo image information of a single-frame radar echo is preprocessed and then input into the three-layer fully connected network model to obtain the image features corresponding to the echo image information of the single-frame radar echo output by the three-layer fully connected network model; the activation function of the three-layer fully connected network model is a non-linear activation function.

[0050] In some embodiments, extracting feature vectors from multiple waypoints to determine the feature vector of the waypoint sequence includes:

[0051] By concatenating the feature vectors of multiple track points, multiple candidate track sequence feature vectors are obtained;

[0052] The feature vector of the candidate track sequence containing the most waypoints is determined as the track sequence feature vector.

[0053] Thirdly, embodiments of this application also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the radar track classification method as described above.

[0054] Fourthly, embodiments of this application also provide a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the radar track classification method as described above.

[0055] Fifthly, embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements the radar track classification method as described above.

[0056] The radar track classification method, apparatus, and storage medium provided in this application generate radar echo images and structured features such as time, position, and velocity from radar echo signals. These features are then processed by a target detection and target tracking network to generate track sequence feature vectors. Radar track classification is performed based on these track sequence feature vectors, which can fully exploit the features of echo data and improve the accuracy of radar track classification and detection. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 This is a flowchart illustrating the radar track classification method provided in the embodiments of this application;

[0059] Figure 2 This is an algorithm structure diagram of the radar track classification method provided in the embodiments of this application;

[0060] Figure 3 This is a schematic diagram of the radar track classification device provided in the embodiments of this application;

[0061] Figure 4 This is a schematic diagram of the physical structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0062] Existing radar track classification research mainly utilizes spectral information, polarization features, and high-resolution range profiles from radar echoes to identify targets within a single frame. However, radar target track identification is strongly correlated with the target's motion information, and there are correlation features between successive tracks. These correlation features are crucial for track identification. Furthermore, radar echo images carry limited information features, necessitating the introduction of structured data such as position and time information for track identification. Therefore, traditional track classification methods have certain limitations: low accuracy, low information utilization, and a lack of effective processing methods for track sequence feature vectors.

[0063] This application proposes a radar track classification method based on multi-feature fusion, aiming to design a track classification method with high accuracy, improve radar track classification efficiency, enhance information utilization efficiency, and meet the radar track classification needs in various scenarios. Traditional methods cannot incorporate track motion information, require complex signal processing, and extract limited information from radar echoes, failing to utilize it effectively. The challenge lies in how to encode motion information into sequence information and how to effectively fuse image and structured features, enabling the track sequence classification network to achieve high detection accuracy with a simple architecture.

[0064] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0065] Figure 1 This is a flowchart illustrating the radar track classification method provided in the embodiments of this application, as shown below. Figure 1 As shown in the embodiment of this application, a method for classifying radar tracks is provided, including:

[0066] Step 101: Acquire radar echo data.

[0067] Specifically, a radar transmitter sends a signal, and a radar receiver receives the reflected echo data to obtain radar echo data. This radar echo data generally includes pure clutter data and target echo data.

[0068] Step 102: Based on the radar echo data, a radar track classification model is used to obtain the radar track classification result. The radar track classification model is obtained by training on the track sequence information obtained through echo image information, echo time information, echo position information and target velocity information, based on the track sequence feature vector and the track category label corresponding to the track sequence feature vector.

[0069] Specifically, the radar echo data is first parsed, converting the binary data containing target echo data into a radar echo image, from which time information and echo position information are extracted. Target detection is then performed on the radar echo image using a target detection algorithm to obtain radar echo target detection boxes. Next, based on a target tracking algorithm, radar echo track sequences are obtained from the radar echo targets, and the target's velocity information is calculated.

[0070] Then, image features of a single-frame radar echo image are obtained based on a feature extraction network model, and the position, velocity, and time information of the target in the single-frame radar echo are fused based on a dynamic multi-layer perceptron (MLP) structure to form a single-frame track point feature vector. The track point feature vector can be represented in the form of a track point feature vector or in other ways; this embodiment does not limit the representation.

[0071] Finally, the feature vectors of multiple track points are concatenated to determine the longest sequence. The longest sequence is then input into a track sequence classification model built on the Transformer structure to obtain the track category. The track sequence classification model outputs the final classification result.

[0072] The radar track classification method provided in this application extracts radar echo images and structured features such as time, position, and velocity from radar echo signals. It then extracts track sequence feature vectors through a target detection and target tracking network and classifies radar tracks based on these feature vectors. This method can fully exploit the features of echo data and improve the accuracy of radar track classification and detection.

[0073] In some embodiments, the training steps of the radar track classification model include:

[0074] Based on the radar echo data, the feature vector of the track point is determined;

[0075] Extract feature vectors from multiple track points to determine the track sequence feature vector;

[0076] The radar track classification model is trained based on the track sequence feature vector and the track category label corresponding to the track sequence feature vector.

[0077] Specifically, the radar track classification model is trained based on known track category labels and acquired track sequence feature vectors. Based on radar echo data, the image features of a single-frame radar echo image and the position, velocity, and time information of the target in that single-frame radar echo are determined, and these are fused using an MLP structure to obtain a single-frame track point feature vector. Multiple track point features are concatenated, and the longest sequence is determined. This longest sequence is input into a track sequence classification model built on a Transformer structure to obtain the track category. The track sequence classification model outputs the final classification result. The radar track classification model is then trained based on the output final classification result and the track category labels to obtain a trained radar track classification model.

[0078] The radar track classification method provided in this application extracts radar echo images and structured features such as time, position, and velocity from radar echo signals. It then extracts track sequence feature vectors through a target detection and target tracking network, and trains a radar track classification model based on the track sequence feature vectors and track classification labels. This method can improve the accuracy of the radar track classification results output by the radar track classification model.

[0079] In some embodiments, determining the feature vector of the track point based on the radar echo data includes:

[0080] Based on the radar echo data, the echo image information, echo time information, echo position information, and target velocity information of a single frame of radar echo are obtained.

[0081] The echo image information, echo time information, echo position information, and target velocity information of the single-frame radar echo are fused to form the track point feature vector.

[0082] Specifically, the radar echo data is parsed, converting binary data containing target echo data into radar echo images, from which time and echo position information is extracted. Target detection is then performed on the radar echo images using a target detection algorithm to obtain radar echo target detection boxes. Next, based on a target tracking algorithm, radar echo track sequences are obtained from the radar echo targets, and the target velocity information is calculated. Finally, image features of a single-frame radar echo image are obtained using a feature extraction network model, and the position, velocity, and time information of the target in each single-frame radar echo are fused using an MLP structure to form a single-frame track point feature vector.

[0083] Preferably, the feature extraction network can first resize the target image to 120×45, then flatten the vector, pass it through a three-layer fully connected network, add a non-linear activation function ReLU, and extract image features. Then, it passes six-dimensional structured features, including time, velocity, and position information, through a single fully connected network to extract the structured data features. The MLP structure can dynamically weight and fused the image features and structured features, with the weights obtained through training, ultimately forming a single-frame track point feature vector.

[0084] The radar track classification method provided in this application extracts radar echo images and structured features such as time, position, and velocity from radar echo signals using a feature extraction network, and then uses an MLP structure to stitch the features together to obtain a single-frame track point feature vector. This method can extract multiple feature information from radar echo signals, thereby improving the accuracy of radar track classification and detection.

[0085] In some embodiments, acquiring echo image information, echo time information, echo position information, and target velocity information of a single-frame radar echo based on the radar echo data includes:

[0086] The radar echo data is analyzed to obtain echo image information, echo time information, and echo location information;

[0087] Target detection is performed on the echo image information, and target velocity information is obtained using a target tracking algorithm.

[0088] Specifically, the radar echo data is parsed, converting binary data containing target echo data into a radar echo image, from which time and echo position information are extracted. Target detection is then performed on the radar echo image using a target detection algorithm to obtain radar echo target detection boxes. Finally, based on a target tracking algorithm, radar echo track sequences are obtained from the radar echo targets, and the target's velocity information is calculated.

[0089] Preferably, the target detection network uses a weight file trained on YOLOv5 and performs target detection on radar echo images based on manual annotation. The target tracking algorithm uses a STARK-based multi-target tracking algorithm to form a complete track sequence image.

[0090] The radar track classification method provided in this application extracts radar echo images and structured features such as time, position, and velocity from radar echo signals. It then extracts track sequence feature vectors through a target detection and target tracking network and classifies radar tracks based on these feature vectors. This method can fully exploit the features of echo data and improve the accuracy of radar track classification and detection.

[0091] In some embodiments, the step of performing target detection on the echo image information and obtaining target velocity information using a target tracking algorithm includes:

[0092] Target detection is performed on the echo image information based on the target detection algorithm to obtain the radar echo target;

[0093] Based on the target tracking algorithm, the trajectory sequence image is obtained from the radar echo target, and the target speed information is calculated.

[0094] Specifically, the radar echo data is parsed, converting binary data containing target echo data into a radar echo image, from which time and echo position information are extracted. Target detection is then performed on the radar echo image using a target detection algorithm to obtain radar echo target detection boxes. Finally, based on a target tracking algorithm, radar echo track sequences are obtained from the radar echo targets, and the target's velocity information is calculated.

[0095] The radar track classification method provided in this application extracts radar echo images and structured features such as time, position, and velocity from radar echo signals. It then extracts track sequence feature vectors through a target detection and target tracking network and classifies radar tracks based on these feature vectors. This method can fully exploit the features of echo data and improve the accuracy of radar track classification and detection.

[0096] In some embodiments, fusing the echo image information, echo time information, echo position information, and target velocity information of the single-frame radar echo to form the track point feature vector includes:

[0097] Image features corresponding to the echo image information of a single-frame radar echo are obtained based on a three-layer fully connected network model, and the structured information corresponding to the echo time information, echo position information and target velocity information of a single-frame radar echo is obtained based on a one-layer fully connected network model.

[0098] By fusing the image features and the features of the structured information, a track point feature vector is obtained.

[0099] Specifically, image features of a single-frame radar echo image are obtained based on a feature extraction network model, and the position, velocity, and time information of the target in the single-frame radar echo are fused based on a dynamic MLP structure to form a single-frame track point feature vector.

[0100] Preferably, the feature extraction network first resizes the target image to 120×45, then flattens the vector, passes it through a three-layer fully connected network, adds the non-linear activation function ReLU, and extracts image features. It then passes six-dimensional structured features, including time, velocity, and position information, through a single fully connected network layer to extract the structured data features. The MLP structure can concatenate image features and structured features, with weights obtained through training, ultimately forming a single-frame track point feature vector.

[0101] The radar track classification method provided in this application extracts radar echo images and structured features such as time, position, and velocity from radar echo signals using a feature extraction network, and then uses an MLP structure to stitch the features together to obtain a single-frame track point feature vector. This method can extract multiple feature information from radar echo signals, thereby improving the accuracy of radar track classification and detection.

[0102] In some embodiments, the step of obtaining the image features corresponding to the echo image information of a single-frame radar echo based on a three-layer fully connected network model includes:

[0103] The echo image information of a single-frame radar echo is preprocessed and then input into the three-layer fully connected network model to obtain the image features corresponding to the echo image information of the single-frame radar echo output by the three-layer fully connected network model; the activation function of the three-layer fully connected network model is a non-linear activation function.

[0104] Specifically, the feature extraction network can first resize the target image to 120×45, then flatten the vector, pass it through a three-layer fully connected network, and add the non-linear activation function ReLU to extract image features.

[0105] The radar track classification method provided in this application extracts radar echo images and structured features such as time, position, and velocity from radar echo signals using a feature extraction network, and then uses an MLP structure to stitch the features together to obtain a single-frame track point feature vector. This method can extract multiple feature information from radar echo signals, thereby improving the accuracy of radar track classification and detection.

[0106] In some embodiments, extracting feature vectors from multiple waypoints to determine the feature vector of the waypoint sequence includes:

[0107] By concatenating the feature vectors of multiple track points, multiple candidate track sequence feature vectors are obtained;

[0108] The feature vector of the candidate track sequence containing the most waypoints is determined as the track sequence feature vector.

[0109] Specifically, multiple track point feature vectors are concatenated to obtain multiple candidate track sequence feature vectors, and the longest candidate track sequence feature vector is determined. The determined track sequence feature vector is then input into the Transformer Block module, which uses a multi-head attention architecture. After vector flattening, a random deactivation layer, a fully connected network, a ReLU nonlinear activation function, and another fully connected network, the radar track classification model outputs the radar track classification result. Generally, the longer the sequence, the higher the detection accuracy.

[0110] The radar track classification method provided in this application extracts radar echo images and structured features such as time, position, and velocity from radar echo signals. It then extracts track sequence feature vectors through a target detection and target tracking network, and outputs radar track classification results based on the track sequence feature vectors through a radar track classification model. This method integrates multiple feature information and extracts sequence features, thereby improving the anti-interference capability of radar track classification and detection, and ultimately improving the accuracy of radar track classification and detection.

[0111] The methods described in the above embodiments will be further illustrated below with specific examples.

[0112] Figure 2 This is an algorithm structure diagram of the radar track classification method provided in the embodiments of this application, as shown below. Figure 2 As shown, the method includes:

[0113] (1) Obtain the original radar echo data, convert the binary data into radar echo images, and parse out the time information and echo position information from them; perform target detection on the radar echo images based on the YOLO algorithm to obtain radar echo target detection boxes; and obtain radar echo track sequence images from radar echo targets based on the STARK tracking algorithm and calculate the target speed information.

[0114] (2) Based on the feature extraction network, the image features of a single-frame radar echo image are obtained. The position information, velocity information and time information of the single-frame radar echo target are fused based on the dynamic MLP structure to form a single-frame track point feature vector.

[0115] (3) Construct a trajectory sequence feature vector classification model based on Transformer, obtain trajectory categories, and output the final classification result.

[0116] (4) The above model is trained based on the graphics processing unit (GPU) to obtain the weight file after training. The model is then deployed to the algorithm server, and the received radar echo signals are processed according to the above process to achieve track classification.

[0117] The dataset inference results show that when the feature vector length of the track sequence is 128, using a two-layer Transformer structure with the attention mechanism of each layer having a head parameter of 2, the track prediction accuracy can reach 84.59% in real-world maritime scenarios for a four-class classification network.

[0118] The radar track classification method provided in this application solves the problem of track sequence classification that traditional methods cannot handle. It also integrates structured features, overcoming the limitation of traditional methods that only use radar echo images. The related algorithms of the radar track classification method provided in this application can be deployed on an algorithm server as needed, without being limited by hardware systems, maximizing the extraction of data features. The radar track classification method provided in this application integrates multi-feature information and extracts sequence features, exhibiting strong anti-interference capabilities, a simple and reliable process, low computational cost, and the ability to achieve efficient and accurate prediction for different categories of tracks.

[0119] Figure 3 This is a schematic diagram of the radar track classification device provided in the embodiments of this application, as shown below. Figure 3 As shown, the radar track classification device provided in this application embodiment includes a first acquisition module 301 and a first processing module 302, wherein:

[0120] The first acquisition module 301 is used to acquire radar echo data;

[0121] The first processing module 302 is used to obtain radar track classification results based on the radar echo data using a radar track classification model. The radar track classification model is obtained by training on track sequence feature vectors and track category labels corresponding to track sequence feature vectors, based on track sequence feature vectors and track sequence feature vectors.

[0122] In some embodiments, the training steps of the radar track classification model include:

[0123] Based on the radar echo data, the feature vector of the track point is determined;

[0124] Extract feature vectors from multiple track points to determine the track sequence feature vector;

[0125] The radar track classification model is trained based on the track sequence feature vector and the track category label corresponding to the track sequence feature vector.

[0126] In some embodiments, determining the feature vector of the track point based on the radar echo data includes:

[0127] Based on the radar echo data, the echo image information, echo time information, echo position information, and target velocity information of a single frame of radar echo are obtained.

[0128] The echo image information, echo time information, echo position information, and target velocity information of the single-frame radar echo are fused to form the track point feature vector.

[0129] In some embodiments, acquiring echo image information, echo time information, echo position information, and target velocity information of a single-frame radar echo based on the radar echo data includes:

[0130] The radar echo data is analyzed to obtain echo image information, echo time information, and echo location information;

[0131] Target detection is performed on the echo image information, and target velocity information is obtained using a target tracking algorithm.

[0132] In some embodiments, the step of performing target detection on the echo image information and obtaining target velocity information using a target tracking algorithm includes:

[0133] Target detection is performed on the echo image information based on the target detection algorithm to obtain the radar echo target;

[0134] Based on the target tracking algorithm, the trajectory sequence image is obtained from the radar echo target, and the target speed information is calculated.

[0135] In some embodiments, fusing the echo image information, echo time information, echo position information, and target velocity information of the single-frame radar echo to form the track point feature vector includes:

[0136] Image features corresponding to the echo image information of a single-frame radar echo are obtained based on a three-layer fully connected network model, and the structured information corresponding to the echo time information, echo position information and target velocity information of a single-frame radar echo is obtained based on a one-layer fully connected network model.

[0137] By fusing the image features and the features of the structured information, a track point feature vector is obtained.

[0138] In some embodiments, the step of obtaining the image features corresponding to the echo image information of a single-frame radar echo based on a three-layer fully connected network model includes:

[0139] The echo image information of a single-frame radar echo is preprocessed and then input into the three-layer fully connected network model to obtain the image features corresponding to the echo image information of the single-frame radar echo output by the three-layer fully connected network model; the activation function of the three-layer fully connected network model is a non-linear activation function.

[0140] In some embodiments, extracting feature vectors from multiple waypoints to determine the feature vector of the waypoint sequence includes:

[0141] By concatenating the feature vectors of multiple track points, multiple candidate track sequence feature vectors are obtained;

[0142] The feature vector of the candidate track sequence containing the most waypoints is determined as the track sequence feature vector.

[0143] Specifically, the radar track classification device provided in this application embodiment can implement all the method steps implemented in the radar track classification method embodiment and achieve the same technical effect. Here, the parts that are the same as those in the method embodiment and the beneficial effects will not be described in detail.

[0144] Figure 4 This is a schematic diagram of the physical structure of the electronic device provided in the embodiments of this application, such as... Figure 4 As shown, the electronic device may include: a processor 410, a communication interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communication interface 420, and the memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions in the memory 430 to execute a radar track classification method, which includes:

[0145] Acquire radar echo data;

[0146] Based on the radar echo data, a radar track classification model is used to obtain the radar track classification result. The radar track classification model is obtained by training on the track sequence information obtained through echo image information, echo time information, echo position information and target velocity information, based on the track sequence feature vector and the track category label corresponding to the track sequence feature vector.

[0147] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0148] In some embodiments, the training steps of the radar track classification model include:

[0149] Based on the radar echo data, the feature vector of the track point is determined;

[0150] Extract feature vectors from multiple track points to determine the track sequence feature vector;

[0151] The radar track classification model is trained based on the track sequence feature vector and the track category label corresponding to the track sequence feature vector.

[0152] In some embodiments, determining the feature vector of the track point based on the radar echo data includes:

[0153] Based on the radar echo data, the echo image information, echo time information, echo position information, and target velocity information of a single frame of radar echo are obtained.

[0154] The echo image information, echo time information, echo position information, and target velocity information of the single-frame radar echo are fused to form the track point feature vector.

[0155] In some embodiments, acquiring echo image information, echo time information, echo position information, and target velocity information of a single-frame radar echo based on the radar echo data includes:

[0156] The radar echo data is analyzed to obtain echo image information, echo time information, and echo location information;

[0157] Target detection is performed on the echo image information, and target velocity information is obtained using a target tracking algorithm.

[0158] In some embodiments, the step of performing target detection on the echo image information and obtaining target velocity information using a target tracking algorithm includes:

[0159] Target detection is performed on the echo image information based on the target detection algorithm to obtain the radar echo target;

[0160] Based on the target tracking algorithm, the trajectory sequence image is obtained from the radar echo target, and the target speed information is calculated.

[0161] In some embodiments, fusing the echo image information, echo time information, echo position information, and target velocity information of the single-frame radar echo to form the track point feature vector includes:

[0162] Image features corresponding to the echo image information of a single-frame radar echo are obtained based on a three-layer fully connected network model, and the structured information corresponding to the echo time information, echo position information and target velocity information of a single-frame radar echo is obtained based on a one-layer fully connected network model.

[0163] By fusing the image features and the features of the structured information, a track point feature vector is obtained.

[0164] In some embodiments, the step of obtaining the image features corresponding to the echo image information of a single-frame radar echo based on a three-layer fully connected network model includes:

[0165] The echo image information of a single-frame radar echo is preprocessed and then input into the three-layer fully connected network model to obtain the image features corresponding to the echo image information of the single-frame radar echo output by the three-layer fully connected network model; the activation function of the three-layer fully connected network model is a non-linear activation function.

[0166] In some embodiments, extracting feature vectors from multiple waypoints to determine the feature vector of the waypoint sequence includes:

[0167] By concatenating the feature vectors of multiple track points, multiple candidate track sequence feature vectors are obtained;

[0168] The feature vector of the candidate track sequence containing the most waypoints is determined as the track sequence feature vector.

[0169] Specifically, the electronic device provided in this application embodiment can implement all the method steps implemented by the method embodiment with the execution subject being an electronic device, and can achieve the same technical effect. Here, the parts that are the same as those in the method embodiment and the beneficial effects will not be described in detail.

[0170] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program that can be stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is capable of executing the radar track classification method provided by the above methods, the method comprising:

[0171] Acquire radar echo data;

[0172] Based on the radar echo data, a radar track classification model is used to obtain the radar track classification result. The radar track classification model is obtained by training on the track sequence information obtained through echo image information, echo time information, echo position information and target velocity information, based on the track sequence feature vector and the track category label corresponding to the track sequence feature vector.

[0173] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a radar track classification method provided by the methods described above, the method comprising:

[0174] Acquire radar echo data;

[0175] Based on the radar echo data, a radar track classification model is used to obtain the radar track classification result. The radar track classification model is obtained by training on the track sequence information obtained through echo image information, echo time information, echo position information and target velocity information, based on the track sequence feature vector and the track category label corresponding to the track sequence feature vector.

[0176] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0177] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, 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 can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0178] It should also be noted that the terms "first," "second," etc., used in the embodiments of this application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of the same class, and the number of objects is not limited. For example, the first object can be one or more.

[0179] In the embodiments of this application, the term "and / or" describes the relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following associated objects have an "or" relationship.

[0180] In the embodiments of this application, the term "multiple" refers to two or more, and other quantifiers are similar.

[0181] In this application, "determining B based on A" means that factor A must be considered when determining B. It is not limited to "B can be determined based solely on A," but should also include: "determining B based on A and C," "determining B based on A, C, and E," "determining C based on A, and further determining B based on C," etc. It can also include using A as a condition for determining B, for example, "when A satisfies the first condition, B is determined using the first method"; or "when A satisfies the second condition, B is determined," or "when A satisfies the third condition, B is determined based on the first parameter," etc. Of course, it can also be a condition where A is a factor in determining B, for example, "when A satisfies the first condition, C is determined using the first method, and B is further determined based on C," etc.

[0182] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for classifying radar tracks, characterized in that, include: Acquire radar echo data; Based on the radar echo data, the radar track classification result is obtained using the radar track classification model; The radar track classification model is trained based on the track sequence feature vector and the track category label corresponding to the track sequence feature vector; The feature vector of the trajectory sequence is obtained in the following way: Based on radar echo data, acquire echo image information, echo time information, echo position information and target velocity information of a single frame of radar echo; By using a multilayer perceptron (MLP) structure, the echo image information, echo time information, echo position information, and target velocity information of the single-frame radar echo are fused to form a track point feature vector. Based on the feature vectors of multiple track points, a track sequence feature vector is obtained; The process of fusing the echo image information, echo time information, echo position information, and target velocity information of the single-frame radar echo to form a track point feature vector includes: Image features corresponding to the echo image information of a single-frame radar echo are obtained based on a three-layer fully connected network model, and the structured information corresponding to the echo time information, echo position information and target velocity information of a single-frame radar echo is obtained based on a one-layer fully connected network model. By fusing the image features and the structured information features, a waypoint feature vector is obtained; The process of obtaining a track sequence feature vector based on multiple track point feature vectors includes: By concatenating the feature vectors of multiple track points, multiple candidate track sequence feature vectors are obtained; The feature vector of the candidate track sequence containing the most waypoints is determined as the track sequence feature vector.

2. The radar track classification method according to claim 1, characterized in that, The step of acquiring echo image information, echo time information, echo position information, and target velocity information of a single frame of radar echo based on the radar echo data includes: The radar echo data is analyzed to obtain echo image information, echo time information, and echo location information; Target detection is performed on the echo image information, and target velocity information is obtained using a target tracking algorithm.

3. The radar track classification method according to claim 2, characterized in that, The step of performing target detection on the echo image information and obtaining target velocity information using a target tracking algorithm includes: Target detection is performed on the echo image information based on the target detection algorithm to obtain the radar echo target; Based on the target tracking algorithm, the trajectory sequence image is obtained from the radar echo target, and the target speed information is calculated.

4. The radar track classification method according to claim 1, characterized in that, The image features corresponding to the echo image information of a single-frame radar echo obtained based on a three-layer fully connected network model include: The echo image information of a single-frame radar echo is preprocessed and then input into the three-layer fully connected network model to obtain the image features corresponding to the echo image information of the single-frame radar echo output by the three-layer fully connected network model; the activation function of the three-layer fully connected network model is a non-linear activation function.

5. A radar track classification device, characterized in that, include: The first acquisition module is used to acquire radar echo data; The first processing module is used to obtain radar track classification results based on the radar echo data using a radar track classification model; the radar track classification model is trained based on track sequence feature vectors and track category labels corresponding to the track sequence feature vectors. The feature vector of the trajectory sequence is obtained in the following way: Based on radar echo data, acquire echo image information, echo time information, echo position information and target velocity information of a single frame of radar echo; By using a multilayer perceptron (MLP) structure, the echo image information, echo time information, echo position information, and target velocity information of the single-frame radar echo are fused to form a track point feature vector. Based on the feature vectors of multiple track points, a track sequence feature vector is obtained; The process of fusing the echo image information, echo time information, echo position information, and target velocity information of the single-frame radar echo to form a track point feature vector includes: Image features corresponding to the echo image information of a single-frame radar echo are obtained based on a three-layer fully connected network model, and the structured information corresponding to the echo time information, echo position information and target velocity information of a single-frame radar echo is obtained based on a one-layer fully connected network model. By fusing the image features and the structured information features, a waypoint feature vector is obtained; The process of obtaining a track sequence feature vector based on multiple track point feature vectors includes: By concatenating the feature vectors of multiple track points, multiple candidate track sequence feature vectors are obtained; The feature vector of the candidate track sequence containing the most waypoints is determined as the track sequence feature vector.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the program, it implements the radar track classification method as described in any one of claims 1 to 4.

Images