Millimeter wave radar detection method, detection device, and storage medium
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CREATOR CHINA TCH CO
- Filing Date
- 2024-11-28
- Publication Date
- 2026-07-21
Smart Images
Figure CN119780854B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of radar technology, and in particular to detection methods, detection equipment and storage media for millimeter-wave radar. Background Technology
[0002] Millimeter-wave radar, as an important sensor technology, is widely used in fields such as autonomous driving, drone navigation, industrial inspection, and environmental monitoring. Because millimeter-wave radar can operate normally in various harsh weather conditions (such as rain, fog, and snow) and can accurately detect the distance, speed, and angle of objects, it has become an indispensable core sensing device for realizing advanced driver assistance systems (ADAS) and autonomous driving systems.
[0003] Conventional millimeter-wave radar detection rate tests are mostly based on manual operation. This often leads to inconsistent statistical data results in complex outdoor scenarios due to different statistical analysis standards used by different testers. As a result, the detection results of millimeter-wave radar have problems of low reliability and low testing efficiency. Summary of the Invention
[0004] The main objective of this application is to provide a detection method, detection equipment, and storage medium for millimeter-wave radar, aiming to solve the technical problem of how to effectively improve the testing reliability and efficiency of millimeter-wave radar.
[0005] To achieve the above objectives, this application proposes a detection method for millimeter-wave radar, which includes: Collect point cloud data from the millimeter-wave radar to be tested and video data from the camera; The millimeter-wave radar and camera to be tested are jointly calibrated to determine the conversion matrix between the millimeter-wave radar and the camera. Detect and label target objects in video data to obtain ground truth data; The point cloud data is projected using a transformation matrix to obtain the correspondence between the point cloud data and the ground truth data; The maximum weight matching algorithm is used to match ground value data and point cloud data with corresponding relationships, and the detection result of the millimeter-wave radar to be detected is determined based on the matching result.
[0006] In one embodiment, the steps of acquiring point cloud data from the millimeter-wave radar to be detected and video data from the camera include: Real-time acquisition of point cloud data from several millimeter-wave radars to be tested, as well as real-time acquisition of video data captured by cameras corresponding to the scene in which the point cloud data occurred; Point cloud data and video data are stored according to timestamps.
[0007] In one embodiment, the step of jointly calibrating the millimeter-wave radar and the camera to be detected, and determining the transformation matrix between the millimeter-wave radar and the camera, includes: The camera's internal parameters are calibrated to obtain the camera's intrinsic parameter matrix; The external parameters of the millimeter-wave radar and camera to be tested are calibrated, and the rotation matrix and translation matrix between the millimeter-wave radar and camera to be tested are obtained by combining the internal parameter matrix. The transformation matrix includes the intrinsic parameter matrix, the rotation matrix, and the translation matrix.
[0008] In one embodiment, the step of detecting and labeling target objects in video data to obtain ground truth data includes: Frame extraction is performed on the video data to obtain several target scene images; The preset deep learning object detection model is used to detect each target scene image and mark the target objects in each target scene image; The system uses a preset algorithm to generate bounding boxes containing each target object, and records the position, category, and confidence score of each target object to obtain detection data. By associating the detection data with timestamps, ground truth data is obtained.
[0009] In one embodiment, the step of projecting point cloud data onto ground truth data using a transformation matrix to obtain the correspondence between the point cloud data and the ground truth data includes: Based on the scene map timestamp corresponding to each target scene map, find the point cloud data corresponding to the scene map timestamp to obtain the target point cloud data; Based on the intrinsic parameter matrix, rotation matrix, and translation matrix, the target point cloud data is projected to obtain the point cloud coordinate position of the target point cloud data on the image coordinate system of the camera. Then, the correspondence between the target point cloud data and the ground truth data is determined based on the point cloud coordinate position.
[0010] In one embodiment, the maximum weight matching algorithm includes a first maximum weight matching algorithm and a second maximum weight matching algorithm. The step of matching ground value data and point cloud data with corresponding relationships using the maximum weight matching algorithm includes: An undirected graph is created using the first maximum weight matching algorithm, and the correspondence between the target point cloud data and the first ground value bounding box of the ground value data is obtained through the undirected graph. Traverse each target point cloud data and determine whether each target point cloud data is within the first truth bounding box corresponding to the target point cloud data; If the target point cloud data is within the first ground truth bounding box corresponding to the target point cloud data, then the target point cloud data within the first ground truth bounding box corresponding to the target point cloud data is determined as the first point cloud data. The first distance between the center of the bounding box of the first point cloud data and the first ground truth bounding box corresponding to the first point cloud data is calculated, and the matching relationship between the first point cloud data and the ground truth data with corresponding relationship is determined based on the first distance. If the target point cloud data is not within the first truth bounding box corresponding to the target point cloud data, then determine whether the target point cloud data is within the second truth bounding box established based on the first truth bounding box corresponding to the target point cloud data. If the target point cloud data is within the second ground truth bounding box corresponding to the target point cloud data, then the target point cloud data within the second ground truth bounding box corresponding to the target point cloud data is determined as the second point cloud data. The second distance between the center of the bounding box of the second point cloud data and the second ground truth bounding box corresponding to the second point cloud data is calculated, and the matching relationship between the second point cloud data and the ground truth data with corresponding relationship is determined based on the second distance.
[0011] In one embodiment, the step of determining the detection result of the millimeter-wave radar to be detected based on the obtained matching result further includes: The test scenario is determined, and lane and distance intervals are divided into the test scenario to obtain multiple lane and distance partitions.
[0012] In one embodiment, the step of determining the detection result of the millimeter-wave radar to be detected based on the obtained matching result includes: Determine the center of the first ground truth bounding box, the center of the second ground truth bounding box, the lane zone and distance zone where each first point cloud data and each second point cloud data are located at each time stamp, and determine the total number of ground truth bounding boxes and the total number of point cloud data for each lane zone and each distance zone at each time stamp. Divide the total number of point cloud data by the total number of ground truth bounding boxes to obtain the detection rate of each lane zone and each distance zone at each timestamp, and determine the detection result based on the detection rate.
[0013] In addition, to achieve the above objectives, this application also proposes a detection device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the millimeter-wave radar detection method as described above.
[0014] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the millimeter-wave radar detection method as described above.
[0015] One or more technical solutions proposed in this application have at least the following technical effects: The detection process involves collecting point cloud data from the millimeter-wave radar under test and video data from the camera; jointly calibrating the millimeter-wave radar and camera to determine the transformation matrix between them; detecting and marking target objects in the video data to obtain ground truth data; projecting the point cloud data using the transformation matrix to obtain the correspondence between the point cloud data and the ground truth data; using a maximum weight matching algorithm to match the ground truth data and point cloud data with the corresponding relationship; and determining the detection result of the millimeter-wave radar under test based on the matching result.
[0016] This application combines the joint calibration results of the camera and millimeter-wave radar (i.e., the transformation matrix) during the data matching process to project point cloud data onto the image coordinate system, obtaining the correspondence between point cloud data and ground truth data, thus achieving spatial alignment. Then, a maximum weight matching algorithm is used to match ground truth data and point cloud data with corresponding relationships, achieving high-precision matching of point cloud data and ground truth data. This improves the reliability of millimeter-wave radar detection results. Since all the above operations are performed on the detection equipment, the low detection efficiency problem of manual operation is avoided. Attached Figure Description
[0017] 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.
[0018] 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.
[0019] Figure 1 A flowchart illustrating the detection method of the millimeter-wave radar in this application (Example 1). Figure 2 A schematic diagram illustrating the continuous acquisition of point cloud data from different millimeter-wave radars to be tested by the testing equipment of this application; Figure 3 This is a schematic diagram illustrating the lane and distance zones in this application; Figure 4 This is a statistical diagram illustrating the detection rate of the ground truth bounding boxes for each lane zone and each distance zone in this application. Figure 5 This is a schematic diagram of the hardware operating environment involved in the detection method of millimeter-wave radar in the embodiments of this application.
[0020] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation
[0021] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.
[0022] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.
[0023] The main solution of this application embodiment is as follows: collecting point cloud data of the millimeter-wave radar to be detected and video data of the camera; jointly calibrating the millimeter-wave radar to be detected and the camera to determine the transformation matrix between the millimeter-wave radar to be detected and the camera; detecting and marking target objects in the video data to obtain ground truth data; projecting the point cloud data through the transformation matrix to obtain the correspondence between the point cloud data and the ground truth data; using the maximum weight matching algorithm to match the ground truth data and point cloud data with the corresponding relationship, and determining the detection result of the millimeter-wave radar to be detected based on the obtained matching result.
[0024] Since conventional millimeter-wave radar detection rate tests are mostly based on manual operation, in complex outdoor scenarios, different statistical analysis standards used by different testers often lead to inconsistent statistical data results, resulting in low reliability and low testing efficiency of millimeter-wave radar detection results.
[0025] This application provides a solution that, during the data matching process, combines the joint calibration results of the camera and millimeter-wave radar, i.e., the transformation matrix, to project point cloud data onto the image coordinate system, obtaining the correspondence between point cloud data and ground truth data, thus achieving spatial alignment. Then, the maximum weight matching algorithm is used to match ground truth data and point cloud data with corresponding relationships, achieving high-precision matching operation between point cloud data and ground truth data, thereby improving the reliability of millimeter-wave radar detection results. Since all the above operations are performed based on the detection equipment, the problem of low detection efficiency caused by manual operation is avoided.
[0026] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device or testing device capable of performing the above functions. The following description uses a testing device as an example to illustrate this embodiment and the subsequent embodiments.
[0027] Based on this, embodiments of this application provide a detection method for millimeter-wave radar, referring to... Figure 1 and combined Figure 2 , Figure 1This is a flowchart illustrating the first embodiment of the millimeter-wave radar detection method of this application.
[0028] In this embodiment, the detection method of the millimeter-wave radar includes steps S10 to S50: Step S10: Collect point cloud data of the millimeter-wave radar to be tested and video data from the camera.
[0029] After establishing the correct connection between the detection equipment and the millimeter-wave radar and camera under test, the detection equipment can acquire point cloud data from the millimeter-wave radar and video data from the camera in real time, as described below: In one feasible implementation, step S10 includes step S11: Step S11: Real-time acquisition of point cloud data of several millimeter-wave radars to be detected, and real-time acquisition of video data captured by the camera corresponding to the scene in which the point cloud data occurred. The point cloud data and video data are stored according to timestamps.
[0030] After establishing correct connections between the testing equipment and the millimeter-wave radar and camera under test, and ensuring that both the radar and camera are functioning normally and synchronously, the testing equipment continuously collects point cloud data from different millimeter-wave radars under test. For details, please refer to... Figure 2 As shown, the green and red point cloud data can be regarded as point cloud data detected by different millimeter-wave radars. At the same time, video data of the corresponding scene is recorded simultaneously. All collected point cloud data and video data are stored according to the corresponding timestamps.
[0031] The timestamp is used for time alignment of point cloud data and video data.
[0032] Step S20: Perform joint calibration of the millimeter-wave radar and camera to be tested, and determine the transformation matrix between the millimeter-wave radar and camera.
[0033] After pre-calibrating the camera's intrinsic parameters, the extrinsic parameters are calibrated using a self-developed joint calibration tool for millimeter-wave radar and camera, resulting in rotation and translation matrices (i.e., transformation matrices). This achieves spatial alignment between the camera and the millimeter-wave radar point cloud, providing a reliable foundation for matching real-world point cloud data. The details are as follows: In one feasible implementation, step S20 includes steps S21-S22: Step S21: The internal parameters of the camera are calibrated to obtain the camera's intrinsic parameter matrix.
[0034] Step S22: The external parameters of the millimeter-wave radar and camera to be detected are calibrated, and the rotation and translation matrices between the millimeter-wave radar and camera are obtained by combining the intrinsic parameter matrix. The transformation matrix includes the intrinsic parameter matrix, the rotation matrix, and the translation matrix.
[0035] Assume the coordinates of the millimeter-wave radar to be detected in the camera coordinate system are as follows: , , It should be noted that, Coordinate system (radar coordinate system) in The pose in the coordinate system (camera coordinate system) is ,in These are pitch, yaw, and roll, respectively. The rotation matrix of the axis is ,exist The rotation matrix of the axis is ,exist The rotation matrix of the axis is , can obtain the right front day Before the coordinate system rotates to the right The coordinate system and the specific intrinsic parameter matrix are shown in Formula 1: ————Formula 1 The specific rotation matrix is shown in Formula 2: ————Formula 2 The specific translation matrix is shown in Formula 3: ——Formula 3 in, , and Here, represents the point cloud position coordinates in the radar coordinate system, K is the camera's intrinsic parameter matrix, and u and v are the point cloud position coordinates in the image coordinate system. , and These are the point cloud position coordinates in the camera coordinate system.
[0036] Step S30: Detect and label the target objects in the video data to obtain ground truth data.
[0037] It should be noted that this embodiment combines deep learning to detect and label target objects in video data, making the generated ground truth data closer to the real value, thereby improving the detection accuracy of the millimeter-wave radar to be detected.
[0038] In one feasible implementation, step S30 includes steps S31 to S35: Step S31: Perform frame extraction on the video data to obtain several target scene images.
[0039] Step S32: Use a preset deep learning object detection model to detect each target scene image and mark the target objects in each target scene image.
[0040] Step S33: Generate bounding boxes containing each target object using a preset algorithm, and record the position, category, and confidence score of each target object to obtain detection data.
[0041] Step S34: Associate the detection data with the timestamp to obtain the ground truth data.
[0042] First, the video data undergoes frame extraction to extract representative scene images (i.e., target scene images) from the real-time video stream. This ensures that the selected target scene images cover different time periods and that the scene targets are in motion. Then, a pre-trained YOLOv8 deep learning object detection model (i.e., the default deep learning object detection model) is used to detect these target scene images. The model analyzes the features of each target scene image, automatically identifying and labeling target objects, including predetermined target categories such as cars, trucks, and buses. During detection, the YOLOv8 algorithm in the YOLOv8 deep learning object detection model generates bounding boxes containing the target objects, recording the position, category, and confidence score of each object, and using this as the detection result, ensuring the accuracy and reliability of the detection results.
[0043] In addition, all detection results will be associated with timestamp information to form a structured dataset. The detection results will be used as ground truth data to match and statistically analyze the point cloud data in subsequent steps. This process not only improves the efficiency of scene analysis, but also lays the foundation for accurate alignment and statistical analysis of subsequent data.
[0044] Step S40: Project the point cloud data using a transformation matrix to obtain the correspondence between the point cloud data and the ground truth data.
[0045] In one feasible implementation, step S40 includes steps S41 to S42: Step S41: Based on the scene map timestamp corresponding to each target scene map, find the point cloud data corresponding to the scene map timestamp on the timestamp to obtain the target point cloud data.
[0046] Step S42: Based on the intrinsic parameter matrix, rotation matrix, and translation matrix, project the target point cloud data to obtain the point cloud coordinate position of the target point cloud data on the image coordinate system of the camera. Then, determine the correspondence between the target point cloud data and the ground truth data based on the point cloud coordinate position.
[0047] First, based on the scene graph timestamp corresponding to the target scene graph, find the timestamp corresponding to the point cloud data that is closest to or corresponds to the scene graph timestamp, ensuring that the two timestamps are as consistent as possible in time, and then determine the found point cloud data as the target point cloud data.
[0048] The determined target point cloud data will be transformed to the camera's image coordinate system to achieve spatial alignment. Therefore, the target point cloud data needs to be projected based on the intrinsic parameter matrix, rotation matrix, and translation matrix. Projection calculations are performed on each point cloud data to calculate the point cloud coordinate position of each target point cloud data in the image coordinate system, thereby determining the correspondence between the target point cloud data and the ground truth data.
[0049] Step S50: The maximum weight matching algorithm is used to match the ground value data and point cloud data that have corresponding relationships, and the detection result of the millimeter-wave radar to be detected is determined based on the obtained matching result.
[0050] In this embodiment, a maximum weighted matching algorithm is used to match the ground truth data calculated in step S40 with the target point cloud data. This algorithm effectively reduces erroneous matching between the ground truth data and the target point cloud data, improving the accuracy and precision of the matching. This, in turn, enhances the detection accuracy of the millimeter-wave radar detection results.
[0051] The entire matching process is divided into a first maximum weight matching process and a second maximum weight matching process. The difference between these two maximum weight matching processes lies in the different rules for adding the truth bounding boxes for the truth data, as detailed below: In one feasible implementation, step S50 includes steps S51 to S55: Step S51: Create an undirected graph using the first maximum weight matching algorithm, and obtain the correspondence between the target point cloud data and the first ground truth bounding box of the ground truth data through the undirected graph; Step S52: Traverse each target point cloud data and determine whether each target point cloud data is within the first true value bounding box corresponding to the target point cloud data. Step S53: If the target point cloud data is within the first ground truth bounding box corresponding to the target point cloud data, then the target point cloud data within the first ground truth bounding box corresponding to the target point cloud data is determined as the first point cloud data. The first distance between the center of the bounding box of the first point cloud data and the first ground truth bounding box corresponding to the first point cloud data is calculated. Based on the first distance, the matching relationship between the first point cloud data and the ground truth data with corresponding relationship is determined.
[0052] In the first maximum weighted matching process, an undirected graph is first created to represent the relationship between the first ground truth bounding boxes and the target point cloud data. Each first ground truth bounding box is traversed, and a node is created for each ground truth bounding box, named gt_x, where x is the index of the ground truth data.
[0053] Iterate through all target point cloud data. For each target point cloud data, check whether its pixel point cloud data coordinates in the pixel coordinate system are within the first ground truth bounding box of a certain ground truth data.
[0054] If a target point cloud data point is within the first ground truth bounding box of a ground truth data point, then the target point cloud data point within the first ground truth bounding box is defined as the first point cloud data point. The first distance between the points of the first point cloud data point and the center of the first ground truth bounding box is calculated, and a first score is calculated based on this first distance. The formula for calculating the score is shown in Formula 4.
[0055] ————Formula 4 It should be noted that score1 is the first score calculated based on the first distance, and distance1 is the first distance.
[0056] If a target point cloud data is not within the first ground truth bounding box corresponding to any ground truth data, then the first score of the target point cloud data is directly determined as none.
[0057] After calculating all the target point cloud data, for the target point cloud data whose first score is not none, add an edge to the undirected graph from the point of the first point cloud data to the center of the bounding box of the first ground truth bounding box. The weight of this edge is the first score corresponding to the first point cloud data.
[0058] Step S54: If the target point cloud data is not within the first truth bounding box corresponding to the target point cloud data, then determine whether the target point cloud data is within the second truth bounding box established based on the first truth bounding box corresponding to the target point cloud data.
[0059] Step S55: If the target point cloud data is within the second ground truth bounding box corresponding to the target point cloud data, then the target point cloud data within the second ground truth bounding box corresponding to the target point cloud data is determined as the second point cloud data. The second distance between the center of the bounding box of the second point cloud data and the second ground truth bounding box corresponding to the second point cloud data is calculated. Based on the second distance, the matching relationship between the second point cloud data and the ground truth data with corresponding relationship is determined.
[0060] Based on the target point cloud data that was determined to have a first score of none in step S52, i.e., the target point cloud data is not within the first ground truth bounding box corresponding to the target point cloud data, a second maximum weight matching calculation is performed on the target point cloud data. Specifically, a second ground truth bounding box is established for each ground truth data with the radius of three-quarters of the diagonal length of the first ground truth bounding box of the ground truth data and the center of the first ground truth bounding box as the center. It is then determined whether the target point cloud data with a first score of none exists within the second ground truth bounding box corresponding to any ground truth data.
[0061] If the target point cloud data exists within any second ground truth bounding box, then the target point cloud data within the second ground truth bounding box is defined as the second point cloud data. The second distance between the points of the second point cloud data and the center of the second ground truth bounding box is calculated, and a second score is calculated based on the second distance. The calculation formula for the score is shown in Formula 5.
[0062] ————Formula 5 It should be noted that score2 is the second score calculated based on the second distance, and distance2 is the second distance.
[0063] If a target point cloud data is not within the second ground truth bounding box corresponding to any ground truth data, then the second score of the target point cloud data is directly determined as none.
[0064] After calculating all the target point cloud data, for the second point cloud data whose second score is not none, add an edge to the undirected graph from the point of the second point cloud data to the center of the bounding box of the second ground truth bounding box. The weight of this edge is the second score corresponding to the second point cloud data.
[0065] As shown above, the first and second maximum weight matching processes determine the matching relationship between the ground truth data and the target point cloud data, ensuring high-quality pairing results in subsequent analysis and statistics. This process lays the foundation for subsequent statistical analysis of the matching results.
[0066] Furthermore, step S60 is included before step S50: Step S60: Determine the test scenario and divide the test scenario into lanes and distance intervals to obtain multiple lane partitions and distance partitions.
[0067] When performing statistical analysis of the matching results, a semi-automatic statistical analysis tool is first used to divide the test scene in detail. Specifically, based on the test scene captured by the camera, multiple lane zones are divided, and two horizontal dividing lines are set to divide the test scene into three distance zones: near, medium, and far. (See example...) Figure 3 As shown, the green quadrilaterals represent lane divisions, and the yellow lines represent distance division lines. By setting lane and distance divisions, the ground truth bounding boxes at each timestamp can be accurately classified.
[0068] Furthermore, step S50 also includes steps S56-S57: Step S56: Determine the center of the first ground truth bounding box, the center of the second ground truth bounding box, the lane partition and distance partition where each first point cloud data and each second point cloud data are located at each time stamp, and determine the total number of ground truth bounding boxes and the total number of point cloud data for each lane partition and each distance partition at each time stamp.
[0069] Step S57: Divide the total number of point cloud data by the total number of ground truth bounding boxes to obtain the detection rate of each lane zone and each distance zone at each timestamp, and determine the detection result based on the detection rate.
[0070] When statistically analyzing data at a specific timestamp, we first iterate through all the first and second ground truth bounding boxes at that timestamp, and then perform classification analysis based on their bounding box centers.
[0071] The classification process is performed in two stages: the first stage is based on lane partitioning. The detection device determines which lane partition a bounding box belongs to based on the positions of the centers of the first and second ground truth bounding boxes. Each lane partition has a corresponding quadrilateral bounding box, defined by the coordinates of four vertices in the image coordinate system. A ray casting method is used to check whether the center of each ground truth bounding box falls within these lane partition bounding boxes. The detection device assigns the first point cloud data and the second point cloud data that match the ground truth bounding boxes to the lane partition where the ground truth bounding box is located. This classification process helps distinguish the lane positions of target objects in the scene image, providing support for further statistical analysis.
[0072] The second classification is based on distance-based partitioning. The detection device determines whether a bounding box belongs to a near, medium, or far-range region based on the position of the centers of the first and second ground truth bounding boxes relative to two defined horizontal dividing lines. Each distance interval is defined according to the actual road scene; for example, near distance might correspond to a closer area in front of the camera's view, while far distance covers areas further away. This step provides statistical information about the distance distribution of target objects, helping to evaluate the detection accuracy of millimeter-wave radar and cameras at different distances.
[0073] After classification, the detection rate is calculated for the ground truth bounding boxes of each lane zone and each distance zone. Taking the detection rate of the first lane zone as an example: based on the classification results, the total number N of ground truth bounding boxes for lane zone 1 is obtained. Based on the matching results of step S40, the first point cloud data and the second point cloud data matched for each ground truth bounding box belonging to lane zone 1 are obtained. Let the total number of matched point cloud data (i.e., the sum of the first point cloud data and the second point cloud data) be C, then the detection rate... Similarly, statistics are performed for each lane zone and each distance zone for each timestamp. The statistical results are stored on the detection equipment for subsequent analysis of the test results of the millimeter-wave radar under test. The statistical results are as follows: Figure 4 As shown.
[0074] Figure 4 The frames in the figure represent timestamps. As shown in the figure, when the timestamp is 4166, there are 6 target point cloud data in lane 1 (i.e., lane section 1), 5 target point cloud data in lane 2 (i.e., lane section 2), 2 target point cloud data in lane 3 (i.e., lane section 3), 2 target point cloud data in lane 4 (i.e., lane section 4), 1 target point cloud data in lane 5 (i.e., lane section 5), and 5 target point cloud data in lane 6 (i.e., lane section 6). There are 15 target point cloud data in the near range, 4 target point cloud data in the medium range, and 2 target point cloud data in the far range. The detection rates on consecutive frames are calculated based on the number of true values. The detection rates for lane 1 are 0.92, lane 2 is 1, lane 3 is 1, lane 4 is 1, lane 5 is 0.97, lane 6 is 1, the detection rate for near distance is 0.99, the detection rate for medium distance is 0.97, and the detection rate for far distance is 0.97.
[0075] In this embodiment, during the data matching process, the point cloud data is projected onto the image coordinate system by combining the joint calibration results of the camera and the millimeter-wave radar, i.e., the transformation matrix, to obtain the correspondence between the point cloud data and the ground truth data, thus achieving spatial alignment. Then, the maximum weight matching algorithm is used to match the ground truth data and point cloud data with corresponding relationships, thereby achieving high-precision matching operation between point cloud data and ground truth data. This improves the reliability of the millimeter-wave radar detection results. Since all the above operations are performed based on the detection equipment, the problem of low detection efficiency caused by manual operation is avoided.
[0076] This application provides a detection device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the millimeter-wave radar detection method in the first embodiment described above.
[0077] The following is for reference. Figure 5 The diagram illustrates a structural schematic of a detection device suitable for implementing embodiments of this application. The detection device in these embodiments may include, but is not limited to, mobile terminals such as laptops, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 5 The detection device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of this application.
[0078] like Figure 5 As shown, the detection device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the detection device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. Communication device 1009 allows the detection device to communicate wirelessly or wiredly with other devices to exchange data. Although the figures show detection devices with various systems, it should be understood that implementing or having all of the systems shown is not required. More or fewer systems may be implemented alternatively.
[0079] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.
[0080] The testing equipment provided in this application, employing the millimeter-wave radar testing method described in the above embodiments, can solve the technical problem of how to effectively improve the testing reliability and efficiency of millimeter-wave radar. Compared with the prior art, the beneficial effects of the testing equipment provided in this application are the same as those of the millimeter-wave radar testing method provided in the above embodiments, and other technical features of this testing equipment are the same as those disclosed in the previous embodiment method, and will not be repeated here.
[0081] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.
[0082] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
[0083] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the millimeter-wave radar detection method in the above embodiments.
[0084] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.
[0085] The aforementioned computer-readable storage medium may be included in the testing equipment; or it may exist independently and not assembled into the testing equipment.
[0086] The aforementioned computer-readable storage medium carries one or more programs. When the aforementioned one or more programs are executed by the detection device, the detection device causes the following to occur: it acquires point cloud data of the millimeter-wave radar to be detected and video data from the camera; it performs joint calibration of the millimeter-wave radar to be detected and the camera to determine the transformation matrix between the millimeter-wave radar to be detected and the camera; it detects and marks target objects in the video data to obtain ground truth data; it projects the point cloud data using the transformation matrix to obtain the correspondence between the point cloud data and the ground truth data; it uses a maximum weight matching algorithm to match the ground truth data and point cloud data with the corresponding relationship, and determines the detection result of the millimeter-wave radar to be detected based on the obtained matching result.
[0087] Computer program code for performing the operations of this application can be written in one or more programming languages or a combination thereof, including object-oriented programming languages such as Java, Smalltalk, and C++, and conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).
[0088] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.
[0089] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.
[0090] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described millimeter-wave radar detection method, thereby solving the technical problem of how to effectively improve the testing reliability and efficiency of millimeter-wave radar. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the millimeter-wave radar detection method provided in the above embodiments, and will not be repeated here.
[0091] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.
Claims
1. A detection method for millimeter-wave radar, characterized in that, The detection method of the millimeter-wave radar includes: Collect point cloud data from the millimeter-wave radar to be tested and video data from the camera; The millimeter-wave radar to be detected and the camera are jointly calibrated to determine the transformation matrix between the millimeter-wave radar to be detected and the camera. The target objects in the video data are detected and marked to obtain ground truth data; The point cloud data is projected using the transformation matrix to obtain the correspondence between the point cloud data and the ground truth data; The maximum weight matching algorithm is used to match ground value data and point cloud data with corresponding relationships, and the detection result of the millimeter-wave radar to be detected is determined based on the obtained matching result. The maximum weight matching algorithm includes a first maximum weight matching algorithm. The step of matching ground value data and point cloud data with corresponding relationships using the maximum weight matching algorithm includes: An undirected graph is created using the first maximum weight matching algorithm, and the correspondence between the target point cloud data and the first ground value bounding box of the ground value data is obtained through the undirected graph. Traverse each of the target point cloud data and determine whether each of the target point cloud data is within the first truth bounding box corresponding to the target point cloud data; If the target point cloud data is within the first truth bounding box corresponding to the target point cloud data, then the target point cloud data within the first truth bounding box corresponding to the target point cloud data is determined as the first point cloud data. The first distance between the center of the bounding box of the first point cloud data and the first truth bounding box corresponding to the first point cloud data is calculated. Based on the first distance, the matching relationship between the first point cloud data and the truth data with corresponding relationship is determined. If the target point cloud data is not within the first truth bounding box corresponding to the target point cloud data, then determine whether the target point cloud data is within the second truth bounding box established based on the first truth bounding box corresponding to the target point cloud data. If the target point cloud data is within the second ground truth bounding box corresponding to the target point cloud data, then the target point cloud data within the second ground truth bounding box corresponding to the target point cloud data is determined as the second point cloud data. The second distance between the second point cloud data and the center of the bounding box of the second ground truth bounding box corresponding to the second point cloud data is calculated. Based on the second distance, the matching relationship between the second point cloud data and the ground truth data with corresponding relationship is determined. The second truth bounding box is a second truth bounding box for each truth data point, with a radius equal to three-quarters of the diagonal length of the first truth bounding box and the center of the first truth bounding box as the center.
2. The detection method of millimeter-wave radar as described in claim 1, characterized in that, The steps for acquiring point cloud data from the millimeter-wave radar to be tested and video data from the camera include: The system acquires point cloud data of several millimeter-wave radars under test in real time, and also acquires video data captured by the camera that corresponds to the scene in which the point cloud data occurred in real time. The point cloud data and the video data are stored according to timestamps.
3. The detection method of millimeter-wave radar as described in claim 2, characterized in that, The step of jointly calibrating the millimeter-wave radar to be detected and the camera to determine the transformation matrix between the millimeter-wave radar to be detected and the camera includes: The internal parameters of the camera are calibrated to obtain the camera's intrinsic parameter matrix; The external parameters of the millimeter-wave radar to be detected and the camera are calibrated, and the rotation matrix and translation matrix between the millimeter-wave radar to be detected and the camera are obtained by combining the internal parameter matrix. The transformation matrix includes the intrinsic parameter matrix, the rotation matrix, and the translation matrix.
4. The detection method of millimeter-wave radar as described in claim 3, characterized in that, The step of detecting and marking target objects in the video data to obtain ground truth data includes: The video data is subjected to frame extraction to obtain several target scene images; The target scene images are detected using a pre-defined deep learning target detection model, and the target objects in each target scene image are marked. A bounding box containing each of the target objects is generated using a preset algorithm, and the position, category, and confidence score of each target object are recorded to obtain detection data; The detection data is associated with the timestamp to obtain the truth data.
5. The detection method of millimeter-wave radar as described in claim 4, characterized in that, The step of projecting the point cloud data using the transformation matrix to obtain the correspondence between the point cloud data and the ground truth data includes: Based on the scene map timestamp corresponding to each target scene map, the point cloud data corresponding to the scene map timestamp is searched on the timestamp to obtain the target point cloud data; Based on the intrinsic parameter matrix, the rotation matrix, and the translation matrix, the target point cloud data is projected to obtain the point cloud coordinate position of the target point cloud data on the image coordinate system of the camera. Then, the correspondence between the target point cloud data and the ground truth data is determined according to the point cloud coordinate position.
6. The detection method of millimeter-wave radar as described in claim 2, characterized in that, Before the step of determining the detection result of the millimeter-wave radar to be detected based on the obtained matching result, the method further includes: The test scenario is determined, and lanes and distance intervals are divided into the test scenario to obtain multiple lane partitions and distance partitions.
7. The detection method of millimeter-wave radar as described in claim 6, characterized in that, The step of determining the detection result of the millimeter-wave radar to be detected based on the obtained matching result includes: Determine the center of the first truth bounding box, the center of the second truth bounding box, the lane partition and distance partition where each of the first point cloud data and each of the second point cloud data is located at each timestamp, and determine the total number of truth bounding boxes and the total number of point cloud data in each lane partition and each of the distance partitions at each timestamp; Divide the total number of point cloud data by the total number of ground truth bounding boxes to obtain the detection rate of each lane partition and each distance partition at each timestamp, and determine the detection result based on the detection rate.
8. A testing device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the detection method of the millimeter-wave radar as claimed in any one of claims 1 to 7.
9. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the detection method of the millimeter-wave radar as described in any one of claims 1 to 7.