Radar performance detection method and device, electronic equipment, computer readable storage medium and vehicle

Abstract

The invention relates to a radar performance detection method and device, electronic equipment, a computer readable storage medium and a vehicle, and the method comprises the steps: determining point cloud data scanned by a radar, and obtaining at least one kind of environment data synchronized with the point cloud data; carrying out detection based on the point cloud data and / or at least one kind of environmental data, and determining at least one of the target object stability, the point loss condition and the environmental detection result scanned by the radar; and determining the degradation condition of the radar performance according to at least one of the target object stability, the point loss condition and the environment detection result. According to the invention, at least one analysis of target object stability, point loss condition and environment detection is carried out on the point cloud data obtained by radar scanning and / or the environment data of radar working, and the degradation condition of radar performance is detected in time, so that targeted processing is carried out; therefore, the real-time performance of radar performance detection and the stability of radar performance are improved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving, and in particular to a radar performance detection method, device, electronic equipment, computer-readable storage medium and vehicle. Background Art

[0002] LiDAR plays a vital role in autonomous driving, and its performance and application are directly related to the safety and reliability of the autonomous driving system. However, LiDAR still faces a series of challenges in practical applications. A common problem is the decline in detection performance in bad weather, such as reduced detection accuracy and reliability, which affects the stability and experience of LiDAR functions. Therefore, it is very important to detect the decline in LiDAR performance in time during the use of the radar so as to carry out targeted processing. Summary of the invention

[0003] One of the objects of the present invention is to provide a radar performance detection method to timely discover radar performance degradation and improve the reliability of radar use; the second object is to provide a radar performance detection device; the third object is to provide an electronic device; the fourth object is to provide a computer-readable storage medium; and the fifth object is to provide a vehicle.

[0004] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0005] A radar performance detection method, comprising:

[0006] Determine point cloud data scanned by the radar, and obtain at least one environmental data synchronized with the point cloud data;

[0007] Perform detection based on the point cloud data and / or the at least one environmental data to determine at least one of the stability of the target object scanned by the radar, the point loss situation, and the environmental detection result;

[0008] Determine the radar performance degradation condition according to at least one of the target object stability, the point loss condition, and the environment detection result.

[0009] According to the above technical means, at least one of the following analyses is performed on the point cloud data obtained by radar scanning and / or the environmental data of the radar operation, including target object stability, point loss, and environmental detection, so that the degradation of radar performance can be detected in time for targeted processing, thereby improving the real-time performance of radar performance detection and the stability of radar performance.

[0010] Further, the determining of the point cloud data scanned by the radar includes:

[0011] Acquire at least one original point cloud scanned by each radar of the plurality of radars;

[0012] Determining a plurality of synchronized original point clouds corresponding to the plurality of radars according to at least one time information corresponding to the at least one original point cloud;

[0013] The multiple synchronized original point clouds are spliced ​​to determine the point cloud data.

[0014] Based on the above technical means, the synchronization and splicing of multiple radar data are achieved, and the accuracy and coverage of detection are improved by using comprehensive radar data.

[0015] Further, determining at least one environmental data synchronized with the point cloud data includes:

[0016] Obtaining at least one sensor data stream collected by at least one sensor;

[0017] For each sensor data stream of at least one sensor data stream, sensor data synchronized with the point cloud data in each sensor data stream is determined according to time information of the point cloud data as environmental data synchronized with the point cloud data, thereby determining the at least one environmental data.

[0018] According to the above technical means, synchronization of radar data and sensor data is achieved, ensuring the accuracy of performance testing based on radar data and sensor data.

[0019] Further, the environmental detection result includes: at least one of weather scene, noise condition, rainfall condition and radar window occlusion condition; the radar working environment detection and / or radar scanning effect detection is performed based on the point cloud data and / or the at least one environmental data to determine at least one of the stability of the target object scanned by the radar, the point loss condition and the environmental detection result, including:

[0020] Performing target object detection based on the point cloud data to determine the stability of the target object;

[0021] And / or, performing ground detection based on the point cloud data to determine the point loss situation;

[0022] and / or, determining the weather scene by performing weather scene detection on the point cloud data;

[0023] And / or, determining the noise condition by performing noise detection on the point cloud data;

[0024] and / or, performing rain detection based on a wiper gear position signal in the at least one environmental data to determine the rain condition;

[0025] And / or, based on the radar window cleanliness signal in the at least one environmental data, the radar window obstruction degree detection is performed to determine the radar window obstruction condition.

[0026] Based on the above technical means, it is possible to analyze the radar performance from the scanning effect of the point cloud data and the radar working environment, thereby improving the accuracy of radar performance detection.

[0027] Further, the determining the stability of the target object by performing target object detection on the point cloud data includes:

[0028] Performing target detection on the point cloud data in the first point cloud data set, determining the total number of target objects corresponding to the first point cloud data set, and the first number of first target objects whose life cycles are less than or equal to a preset life cycle threshold value corresponding to the first point cloud data set; the first point cloud data set includes at least one point cloud data within a first time window;

[0029] determining a first indicator based on the first number and the total number of target objects corresponding to the first point cloud data set;

[0030] and / or,

[0031] determining a current speed based on at least one environmental data;

[0032] In the case where the current speed is greater than or equal to the first speed threshold, determining the total number of target objects corresponding to the second point cloud data set and the second number of second target objects in the target motion direction corresponding to the second point cloud data set and having a speed less than the second speed threshold by performing target detection on the point cloud data in the second point cloud data set; the second point cloud data set includes at least one point cloud data within the second time window;

[0033] determining a second indicator based on the second number and the total number of target objects corresponding to the second point cloud data set;

[0034] The target object stability is determined based on at least one of the first indicator and the second indicator.

[0035] Further, the first indicator represents a ratio of the first number to the total number of target objects corresponding to the first point cloud data set; the second indicator represents a ratio of the second number to the total number of target objects corresponding to the second point cloud data set, and the determining of the target object stability based on at least one of the first indicator and the second indicator includes:

[0036] When at least one of the first indicator and the second indicator is greater than or equal to a first ratio threshold, determining that the target object stability is a first value; the first value indicates that the target object stability is extremely poor;

[0037] In the case where the first indicator is less than the first ratio threshold and greater than or equal to the second ratio threshold, and the second indicator is less than the first ratio threshold and greater than or equal to the second ratio threshold, determining that the target object stability is a first value; and the second ratio threshold is less than the first ratio threshold;

[0038] When the first indicator is less than the first ratio threshold and greater than or equal to the second ratio threshold, or when the second indicator is less than the first ratio threshold and greater than or equal to the second ratio threshold, the target object stability is determined to be a second value; the second value indicates that the target object has poor stability.

[0039] According to the above technical means, the stability of the target object is detected in real time through point cloud data, and the degradation of radar performance can be evaluated from the dimension of radar scanning the target object, so as to timely discover the degradation of radar performance so as to take targeted measures, thereby improving the real-time performance of radar detection and the stability of radar functions.

[0040] Further, the determining the point loss situation by performing ground detection on the point cloud data includes:

[0041] Classify and detect the point cloud data through a ground detection network to determine ground points in the point cloud data;

[0042] Determine a target area in the field of view of the point cloud data, and determine a historical average number of sampling points in the target area;

[0043] The number of echo points and the number of ground points of the point cloud data in the target area are counted, and the point loss situation is determined according to the historical average number of sampling points, the number of echo points and the number of ground points.

[0044] Further, determining the point loss situation according to the historical average sampling point number, the echo point number and the ground point number includes:

[0045] When it is determined that the point cloud data meets the preset point loss condition according to the historical average sampling point number, the echo point number and the ground point number, the point cloud data is recorded, and it is determined whether the next point cloud data meets the preset point loss condition;

[0046] If the recorded point cloud data satisfying the preset point loss condition is greater than or equal to a preset first quantity threshold, then determining that the ground point loss situation is a point loss state;

[0047] If the recorded point cloud data that meets the preset point loss condition is less than or equal to the preset second quantity threshold, the current state of the ground point loss situation is determined to be a no-point loss state; the preset second quantity threshold is less than the preset first quantity threshold.

[0048] According to the above technical means, by evaluating the radar's point loss situation through ground point detection, the reduction or degradation of radar performance can be discovered in time for targeted processing, thereby improving the real-time performance of radar detection and the reliability of radar application.

[0049] Further, the determining the noise situation by performing noise detection on the point cloud data includes:

[0050] Performing noise detection on the point cloud data through a noise detection network to determine the number of noise points in the point cloud data;

[0051] The noise situation is determined based on the number of noise points and the total number of points in the point cloud data.

[0052] Further, determining the noise situation based on the number of noise points and the total number of points in the point cloud data includes:

[0053] Determine the noise condition of multiple frames of point cloud data including the point cloud data;

[0054] For the multi-frame point cloud data, when the ratio between the number of noise points and the total number of points is greater than or equal to the noise point ratio threshold, and the number of point cloud data is greater than or equal to a preset third number threshold, the noise situation is determined to be high noise.

[0055] It can be understood that by detecting the proportion of noise points in the point cloud data through the noise detection network, the reduction or degradation of radar performance can be discovered in time for targeted processing, thereby improving the real-time performance of radar detection and the reliability of radar applications.

[0056] Further, the performing of rain detection based on the wiper gear position signal in the at least one environmental data to determine the rain situation includes:

[0057] When the wiper gear position signal indicates automatic gear, obtaining at least one of a wiper speed and a wiper operation time;

[0058] The rainfall condition is determined according to at least one of the wiper speed and the wiper operation time.

[0059] Further, the performing of rain detection based on the wiper gear position signal in the at least one environmental data to determine the rain situation includes:

[0060] When the wiper gear position signal indicates a manual gear, obtaining a manual gear trigger signal and a wiper speed;

[0061] Determining the wiper triggering times according to the manual gear triggering signal;

[0062] The rainfall condition is determined according to the wiper triggering times.

[0063] According to the above technical means, the rainfall conditions in the radar working environment are detected through the wiper gear signal, so that the reduction or degradation of radar performance can be discovered in time for targeted processing, thereby improving the real-time performance detection of radar and the reliability of radar application.

[0064] It can be understood that by detecting the rainfall conditions in the radar working environment through the wiper signal, the reduction or degradation of radar performance can be discovered in time for targeted processing, thereby improving the real-time performance of radar detection and the reliability of radar application.

[0065] Further, the cleanliness signal includes: at least one partition on the radar window corresponds to at least one regional cleanliness signal; the radar window obstruction degree detection based on the radar window cleanliness signal in the at least one environmental data to determine the radar window obstruction condition includes:

[0066] According to the at least one regional cleanliness signal, statistically analyze the regional cleanliness signals within at least one preset regional range on the radar window, and determine at least one regional statistical result corresponding to the at least one preset regional range; the preset regional range includes one or more partitions;

[0067] The radar window obstruction condition is determined according to the at least one regional statistical result.

[0068] Further, the determining the weather scene by performing weather scene detection on the point cloud data includes:

[0069] Predicting the point cloud data through a weather scene classification network to determine a classification result corresponding to the point cloud data;

[0070] Determine the weather scene by counting multiple classification results corresponding to multiple point cloud data

[0071] It can be understood that by detecting the weather scene type of the radar working environment through the weather scene classification network signal, the reduction or degradation of radar performance can be discovered in time for targeted processing, thereby improving the real-time performance detection of radar and the reliability of radar application.

[0072] A radar performance detection device, comprising:

[0073] An acquisition module, used to determine point cloud data scanned by the radar, and to acquire at least one environmental data synchronized with the point cloud data;

[0074] A detection module, configured to perform detection based on the point cloud data and / or the at least one environmental data to determine at least one of stability of a target object scanned by the radar, a point loss situation, and an environmental detection result;

[0075] A determination module is used to determine the degradation of radar performance according to at least one of the target object stability, the point loss situation and the environmental detection result.

[0076] Beneficial effects of the present invention:

[0077] The present invention detects the target object based on the point cloud data scanned by the radar to determine the stability of the target object scanned by the radar, performs point loss detection based on the point cloud data to determine the point loss situation in the point cloud data scanned by the radar; and performs radar working environment detection based on at least one environmental data to determine the environmental detection result. In this way, the degradation of radar performance is determined based on at least one of the stability of the target object, the point loss situation, and the environmental detection. Thus, by performing at least one analysis of the stability of the target object, the point loss situation, and the environmental detection on the point cloud data obtained by the radar scan and / or the environmental data of the radar working, the degradation of radar performance is detected in time so as to perform targeted processing, thereby improving the real-time performance detection of radar and the stability of radar performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 A schematic diagram of an optional flow chart of the radar performance detection method provided by the present invention;

[0079] Figure 2 A schematic diagram of a radar deployment method on a vehicle provided by the present invention;

[0080] Figure 3 A schematic diagram of the synchronization process of point cloud data and sensor signal stream provided by the present invention;

[0081] Figure 4 A schematic diagram of an optional flow chart of the radar performance detection method provided by the present invention;

[0082] Figure 5 A schematic diagram of an optional flow chart of the radar performance detection method provided by the present invention;

[0083] Figure 6 A schematic diagram of the target area provided by the present invention;

[0084] Figure 7 A schematic diagram of the data relationship between ground points, echo points and sampling points provided by the present invention;

[0085] Figure 8 A schematic diagram of the process of detecting point loss provided by the present invention;

[0086] Fig. 9 A schematic diagram of an optional flow chart of the radar performance detection method provided by the present invention;

[0087] Fig.10 A schematic diagram of an optional flow chart of the radar performance detection method provided by the present invention;

[0088] Fig.11 A schematic diagram of an optional flow chart of the radar performance detection method provided by the present invention;

[0089] Fig.12 A schematic diagram of an optional flow chart of the radar performance detection method provided by the present invention;

[0090] Fig.13 A schematic diagram of an optional flow chart of the radar performance detection method provided by the present invention;

[0091] Fig.14 A schematic diagram of a process of applying the radar performance detection method provided by the present invention to an actual scenario;

[0092] Fig.15 A schematic diagram of the structure of a radar performance detection device provided by the present invention;

[0093] Fig.16 This is a schematic structural diagram of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0094] The following will describe the embodiments of the present invention with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention, not for limiting the scope of protection of the present invention.

[0095] LiDAR plays a vital role in autonomous driving, and its performance and application are directly related to the safety and reliability of the autonomous driving system. LiDAR is the abbreviation of laser detection and ranging system. It measures the distance, speed, angle and other information of the target by emitting laser beams and receiving their echoes, and then constructs a three-dimensional model of the target object. LiDAR has the advantages of high precision, high resolution, wide detection range, good light adaptability, etc., and is an important component of the perception layer of autonomous driving. At present, LiDAR sensors have been widely used in autonomous driving environment perception. Through the perception algorithm combining traditional and deep learning, it can realize the recognition of dynamic obstacles in the environment such as vehicles, vehicles, cyclists, etc.; the recognition of static obstacles such as cones, water barriers, tires, cartons, etc.; the recognition of road elements such as lane lines, curbs, zebra crossings, etc.; it provides accurate perception performance for autonomous driving and greatly improves the safety of autonomous driving systems; although LiDAR has broad application prospects in autonomous driving, its application still faces a series of challenges, the most typical of which is the performance degradation in bad weather. The limitations of LiDAR under bad weather conditions are mainly reflected in the detection accuracy and reliability. The scenes that affect the detection accuracy of LiDAR are mainly There are: 1. Heavy rain and blizzard: Under extreme weather conditions such as heavy rain or blizzard, raindrops and snowflakes will scatter and attenuate laser signals, resulting in a shortened detection distance and reduced accuracy of the lidar; in addition, the accumulation of snowflakes may form obstacles, further affecting the scanning range and accuracy of the lidar; although light rain and light snow have relatively little impact on the lidar, it may still have a certain impact on the detection performance of the lidar in the case of long-term or continuous precipitation; 2. Haze: Tiny particles in fog and haze (such as water droplets, dust, etc.) will scatter laser signals, causing the lidar's echo signal to weaken, making it difficult to accurately measure the position and shape of surrounding objects, and will also cause misidentification of the laser perception algorithm, seriously affecting the stability and experience of the function.

[0096] Therefore, this embodiment aims to provide a radar performance detection method to determine the performance degradation of the laser radar algorithm in real time. In this way, when the radar performance degradation is detected in time, the back-end module can use and process the signal output by the laser radar algorithm in a targeted manner according to the performance degradation signal provided by this method, thereby ensuring the robustness and safety of the system.

[0097] The present invention is mainly to solve the stability and safety impact of the automatic driving function caused by the failure of the algorithm or abnormal conditions due to the degradation of the laser radar performance caused by extreme weather (rain, snow, fog, dust), ensure the robustness of the system and timely downgrade the system in the case of system failure, and ensure the safety of driving and the reliability of the function. The present invention accesses the original point cloud information. For example, the original point cloud information can include the three-dimensional coordinates of the point cloud, reflectivity, point cloud confidence (judgment of whether the point cloud is a real obstacle reflection point), etc., and accesses the vehicle's own wiper rain sensor, the window occlusion flag of the laser radar and other signals, and realizes the judgment of the performance degradation of the laser radar algorithm through the combination of traditional and deep learning algorithms, and outputs the judgment signal to the back end. The radar performance detection method provided in the embodiment of the present invention can be applied to electronic devices in a system or device equipped with at least one radar. Among them, the electronic device can be implemented as a terminal. In some embodiments, the system or device equipped with at least one radar can include a mobile tool such as a vehicle, and the electronic device can include a vehicle computer, a vehicle control unit (VCU) or a vehicle-mounted terminal and other terminal devices on the vehicle. Taking the vehicle-mounted terminal as an example, the vehicle may be configured with at least one radar connected to at least one sensor, and the at least one sensor is used to obtain at least one environmental data. The vehicle-mounted terminal obtains at least one point cloud data scanned by the radar and at least one environmental data synchronized with the point cloud data, and executes the radar performance detection method of the embodiment of the present invention. Alternatively, in some embodiments, the electronic device may also include a server. Exemplarily, the point cloud data scanned by the radar and at least one environmental data synchronized with the point cloud data are sent to the server by the vehicle, and the server executes the radar performance detection method of the embodiment of the present invention based on the acquired point cloud data and at least one environmental data.

[0098] An embodiment of the present invention provides a radar performance detection method, such as Figure 1 As shown, including S101-S103, as follows:

[0099] S101. Determine point cloud data scanned by a radar, and obtain at least one type of environmental data synchronized with the point cloud data.

[0100] In S101, point cloud data obtained by radar scanning is obtained, and the point cloud data includes time information of the radar scanning. Exemplarily, the time information may include a timestamp representing the start or end of the radar scanning. And, at least one environmental data synchronized with the time information of the point cloud data is received.

[0101] In some embodiments, the point cloud data includes a point cloud obtained from at least one radar scan.

[0102] In some embodiments, the electronic device receives in real time at least one sensor signal stream detected by at least one sensor on the vehicle or other mobile tool in which it is located. At least one sensor is used to detect the environment in which it is located in real time to obtain sensor signal streams corresponding to each sensor. Among them, at least one sensor and at least one radar are configured on the same device, so the environment in which at least one sensor is located can represent the working environment of at least one radar. The sensor data stream contains sensor data collected by the sensor at multiple time points. According to the time information of the point cloud data, the sensor data synchronized with it can be determined from the sensor data stream as environmental data, so that at least one environmental data synchronized with the point cloud data can be determined from the at least one sensor data stream.

[0103] In some embodiments, the above process of determining the point cloud data scanned by radar may include: obtaining at least one original point cloud scanned by each radar in multiple radars; determining multiple synchronized original point clouds corresponding to multiple radars based on at least one time information corresponding to at least one original point cloud; and splicing multiple synchronized original point clouds to determine the point cloud data.

[0104] For example, Figure 2 As shown, the vehicle is equipped with two left and right laser radars, which scan in real time to obtain at least one original point cloud corresponding to the left laser radar (at least one left radar point cloud) and at least one original point cloud corresponding to the right laser radar (at least one right radar point cloud). Each left radar point cloud in the at least one left radar point cloud corresponds to a left radar point cloud timestamp T l , each right radar point cloud in at least one right radar point cloud corresponds to a right radar point cloud timestamp T r By using at least one left radar point cloud timestamp T l With at least one right radar point cloud timestamp T r Perform time synchronization judgment to determine the time-synchronized left radar point cloud and right radar point cloud. For example, the left radar point cloud timestamp T l As a benchmark, determine the l Alternatively, you can also use the right radar point cloud timestamp T r As a benchmark, determine the r Synchronized left radar point cloud. A frame of left radar point cloud and a frame of right radar point cloud synchronized in time are used as multiple synchronized original point clouds, and point clouds are spliced ​​to form a frame of spliced ​​point cloud as point cloud data. Here, the time information of the point cloud data can be determined based on the time information of the synchronized original point cloud. For example, the left radar point cloud timestamp T l The time information of the point cloud data obtained is based on the left radar point cloud timestamp T l OK. Take the right radar point cloud timestamp Tr The time information of the point cloud data is synchronized and spliced ​​based on the right radar point cloud timestamp T r The specific selection is made according to the actual situation and is not limited in the embodiment of the present invention.

[0105] In some embodiments, the above-mentioned determination of at least one environmental data synchronized with the point cloud data includes: acquiring at least one sensor data stream collected by at least one sensor; for each sensor data stream of the at least one sensor data stream, determining the sensor data synchronized with the point cloud data in each sensor data stream according to the time information of the point cloud data as the environmental data synchronized with the point cloud data, thereby determining at least one environmental data.

[0106] For example, Figure 3 As shown, at least one sensor data stream may include a vehicle signal stream and a wiper gear signal stream. The signal cycle of the vehicle signal stream is as follows: Figure 3 As shown in the vehicle signal cycle axis, at each Tesp moment (such as Figure 3 The whole vehicle signal is reported once at the Tesp1 moment, Tesp2 moment, ... Tesp11 moment in the wiper position signal flow, with a cycle of 10ms, thereby forming a whole vehicle signal flow. For example, the whole vehicle signal may include signals such as vehicle speed. The signal cycle of the wiper gear position signal flow may be as follows: Figure 3 As shown in the wiper gear signal cycle axis, at each Tbcm moment (such as Figure 3 The wiper position signal is reported once at the Tbcm1 moment, Tbcm2 moment, ... Tbcm7 moment in the figure, with a cycle of 20ms, thereby forming a wiper position signal stream. The scanning cycle of the point cloud data can be as follows: Figure 3 As shown in the period axis of the lidar signal, at T I Moment (such as Figure 3 T I1 Moment and T I2 At time T, a radar scan begins. r Moment (such as Figure 3 T r1 Moment and T r2 At the time (1000 ms), a radar scan is completed, and a point cloud data is obtained with a scanning cycle of 22 ms. Among them, the laser radar signal cycle axis can be the signal cycle axis corresponding to the radar used as a reference for point cloud splicing among multiple radars. Figure 3 It can be seen that the vehicle signal and the wiper gear signal (such as Figure 3 The vehicle signal and wiper gear position signal corresponding to the moment in the middle dotted box) are used as environmental data synchronized with the point cloud data.

[0107] Exemplarily, the sensor includes a rain sensor and / or a wiper; the rain sensor and / or the wiper are used to detect the amount of rainfall in the environment. The sensor data stream includes a rain signal stream and / or a wiper gear information stream detected by the rain sensor. According to the information timestamp T of each rain signal and / or wiper gear information in the received rain signal stream and / or wiper gear information stream, bcm , perform time synchronization judgment with the time information of the point cloud data, and use the rainfall signal and / or wiper gear information synchronized with the point cloud data as one of the at least one environmental data.

[0108] Exemplarily, the sensor includes a speed sensor, which receives a vehicle speed information stream collected by a speed sensor configured on the vehicle, and calculates the timestamp T of each vehicle speed information in the vehicle speed information stream. v , perform time synchronization judgment with the time information of the point cloud data, and use the vehicle speed information synchronized with the point cloud data as one of the at least one environmental data.

[0109] In some embodiments, point cloud data and at least one environmental data synchronized with the point cloud data, such as vehicle speed information or wiper gear information, can be added to a data synchronization container queue, and then the point cloud data and at least one environmental data synchronized with the point cloud data can be taken out from the data synchronization container queue for radar performance detection.

[0110] S102: Perform detection based on point cloud data and / or at least one environmental data to determine at least one of stability of the target object scanned by the radar, point loss conditions, and environmental detection results.

[0111] In S102, the radar working environment and / or the radar scanning effect are detected based on the point cloud data and / or at least one environmental data. For example, the target object scanned by the radar is detected, and / or the point loss situation in the point cloud scanned by the radar is detected, and / or the radar working environment is detected, etc., to determine at least one of the stability of the target object scanned by the radar, the point loss situation and the environmental detection result.

[0112] In some embodiments, the environmental detection result includes at least one of a weather scene, a noise condition, a rainfall condition, and a radar window obstruction condition. The process of S102 may include:

[0113] Detect target objects based on point cloud data and determine the stability of target objects;

[0114] and / or, performing ground detection based on point cloud data to determine missing points;

[0115] and / or, determining the weather scene by performing weather scene detection on the point cloud data;

[0116] and / or, determining noise conditions by performing noise detection on the point cloud data;

[0117] and / or, performing rain detection based on a wiper gear position signal in at least one environmental data to determine the rain condition;

[0118] And / or, based on a cleanliness signal of the radar sunshade in at least one environmental data, a window obstruction degree detection is performed to determine the radar window obstruction condition.

[0119] Exemplarily, in the process of detecting target objects based on point cloud data and determining the stability of target objects, the point cloud data can be input into a radar target detection network that has been trained with a large amount of real vehicle target annotation data. Through the radar target detection network, the stability of the target objects scanned by the radar is evaluated, for example, position stability, speed stability, confidence stability, category stability, etc. are evaluated, and the stability of the target object is output.

[0120] For example, in the process of performing ground detection based on point cloud data and determining the missing point situation, the point cloud data, such as the spliced ​​point cloud, can be passed through a pre-trained ground extraction network G net , perform fast local point cloud ground segmentation and classify the point cloud into ground points P g and non-ground point P ng , count the number of ground points in the specified area, form a ground point distribution threshold based on the statistical results, perform ground point loss statistics for each frame of input point cloud data based on the ground point threshold, accumulate the ground point loss rate over multiple frames, and determine the point loss situation of the radar scan based on the multi-frame accumulation results.

[0121] Exemplarily, in the process of detecting weather scenes and determining weather scenes through point cloud data, the point cloud data can be input into a radar scene classification network that has been trained with a large amount of real vehicle scene annotation data. The radar scene classification network performs scene classification judgment based on the point cloud data, and accumulates the classification results over multiple frames to determine the weather scene.

[0122] For example, in the process of determining the noise situation by performing noise detection on point cloud data, the point cloud data can be input into a radar noise classification network trained with a large amount of real vehicle noise annotated data, and the radar noise classification network can classify noise points in various weather scenes including rainy days, snowy days, dust, etc. The noise points of each category are counted, and the noise point statistics are accumulated over multiple frames to determine the noise situation.

[0123] For example, for the process of detecting the amount of rain based on the wiper gear position signal in at least one environmental data and determining the amount of rain, the wiper and rain sensor information and the vehicle speed information can be connected for judgment. According to the rain sensor gear position, the wiper operation cycle, and the vehicle speed, three gear position information of light rain, no rain, and medium to heavy rain are output as inputs for the next stage comprehensive judgment algorithm:

[0124] Exemplarily, the process of detecting the degree of window obstruction based on the cleanliness signal of the radar hood in at least one environmental data to determine the radar window obstruction situation can include statistically analyzing the cleanliness signals of various areas on the radar hood, performing threshold judgment on the cleanliness signals counted for each area, and accumulating multiple frames for the number of frames that exceed the threshold to determine the radar window obstruction situation.

[0125] It should be noted that, according to other environmental factors that affect radar performance in actual applications, corresponding environmental data can be obtained for detection to evaluate radar performance. The specific selection is made according to actual conditions and is not limited in the embodiments of the present invention.

[0126] S103: Determine the degradation of radar performance according to at least one of target object stability, point loss, and environmental detection result.

[0127] In S103, radar performance evaluation is performed based on at least one of the target object stability, point loss situation, and environment detection result obtained in the above process, so as to determine the degradation of radar performance.

[0128] In some embodiments, a fusion judgment module can be used to receive the noise conditions output by the noise detection judgment module, the weather scenes output by the scene classification judgment module, the point loss conditions output by the ground point loss judgment module, the target object stability output by the target stability judgment module, the rainfall conditions output by the rainfall information judgment module, and the radar window obstruction conditions output by the hood dirtiness judgment module, and combine with preset degradation judgment rules to evaluate and judge the degradation of the radar function and output the degradation flag under different conditions.

[0129] It can be understood that the embodiment of the present invention performs target object detection based on the point cloud data scanned by the radar to determine the stability of the target object scanned by the radar, performs point loss detection based on the point cloud data to determine the point loss situation in the point cloud data scanned by the radar; and performs radar working environment detection based on at least one environmental data to determine the environmental detection result. In this way, the degradation of radar performance is determined based on at least one of the stability of the target object, the point loss situation, and the environmental detection. Thereby, by performing at least one analysis of the stability of the target object, the point loss situation, and the environmental detection on the point cloud data obtained by the radar scan and / or the environmental data of the radar working, the degradation of radar performance is detected in time so as to perform targeted processing, thereby improving the real-time performance detection of radar and the stability of radar performance. Furthermore, compared to the current related art in which the LiDAR hardware can only make a single LiDAR performance degradation judgment by judging the echo energy of the point cloud, the embodiment of the present invention integrates the original point cloud information provided by the hardware, the echo energy information, the wiper / rainfall sensor information, and the LiDAR algorithm output information to form an integrated solution combining software and hardware. Compared with the pure hardware solution, the embodiment of the present invention is more robust and has higher detection accuracy. It provides a more reliable LiDAR algorithm performance degradation solution for the back-end module, can reduce the misjudgment caused by the pure hardware method, and provide relatively reliable input for the back-end when the LiDAR hardware performance is degraded in different weather scenarios, thereby enhancing the robustness and reliability of the overall system.

[0130] In some embodiments, Figure 4 As shown, the process of detecting the target object based on the point cloud data and determining the stability of the target object may include:

[0131] S201, performing target detection on the point cloud data in the first point cloud data set, determining the total number of target objects corresponding to the first point cloud data set, and the number of first target objects corresponding to the first point cloud data set whose life cycle is less than or equal to a preset life cycle threshold; determining a first indicator based on the number of first target objects and the total number of target objects corresponding to the first point cloud data set;

[0132] and / or,

[0133] Determine a current speed based on at least one environmental data; when the current speed is greater than or equal to a first speed threshold, determine the total number of target objects corresponding to the second point cloud data set and the number of second target objects in the target motion direction corresponding to the second point cloud data set with a speed less than the second speed threshold by performing target detection on the point cloud data in the second point cloud data set; determine a second indicator based on the number of second target objects and the total number of target objects corresponding to the second point cloud data set.

[0134] In S201, the stability of the target object scanned by the radar may be evaluated based on the target object life cycle output by the target detection network and / or the motion change of the target object.

[0135] Regarding the method of evaluating the stability of the target object according to the life cycle of the target object, the first point cloud data set includes at least one point cloud data within the first time window. That is to say, taking the first time window as the statistical period, one (one frame) of point cloud data is obtained each time through radar scanning, thereby obtaining at least one point cloud data within the first time window as the first point cloud data set. Target detection is performed on each point cloud data in the first point cloud data set to obtain at least one target object in each point cloud data and the life cycle of each target object, so that the target object whose life cycle is less than or equal to the preset life cycle threshold can be determined as the first target object. In the first point cloud data set, the total number of target objects is counted as the total number of target objects corresponding to the first point cloud data set, and the total number of first target objects is counted as the first number.

[0136] In this way, the ratio of the first target objects whose life cycle is less than or equal to the preset life cycle threshold to the total number of target objects corresponding to the first point cloud data set and the first number can be determined to determine the first indicator.

[0137] In some embodiments, the first indicator may be determined based on a ratio of a total number of target objects corresponding to the first point cloud data set to the first number.

[0138] For the method of evaluating the stability of the target object according to the change of the movement of the target object, the current speed can be determined according to at least one environmental data, such as the current speed of the vehicle collected by the speed sensor. Here, the current speed represents the displacement speed of the radar. When the current speed is greater than or equal to the first speed threshold, statistics are performed in the second time window, and at least one point cloud data determined by radar scanning in the second time window is used as the second point cloud data set. Here, the second time window and the first time window can be the same or different. Through the target detection network, target detection is performed on each point cloud data in the second point cloud data set, and the target detection network can output the speed information of each target object in each point cloud data, so that the target object with a speed less than or equal to the second speed threshold in the target motion direction, such as the vehicle driving directly in front of the vehicle, can be determined as the second target object. Among them, the second speed threshold is less than or equal to the first speed threshold. In the second point cloud data set, the total number of target objects is counted as the total number of target objects corresponding to the second point cloud data set, and the total number of second target objects is counted as the second number.

[0139] In this way, when the displacement speed of the radar is greater than or equal to the first speed threshold, the second target object whose moving speed is less than the second speed or equal to the second speed threshold, that is, the target object with abnormal movement that obviously does not match the displacement speed of the radar, can be determined. According to the total number of target objects within the statistical period (second time window) and the total number of second target objects (second number), the proportion of the target objects with abnormal movement to the total number of target objects can be determined, thereby determining the second indicator.

[0140] In some embodiments, the second indicator may be determined based on a ratio of the total number of target objects corresponding to the second point cloud data set to the second number.

[0141] S202: Determine the stability of the target object based on at least one of the first indicator and the second indicator.

[0142] In some embodiments, when at least one of the first indicator and the second indicator is greater than or equal to the first ratio threshold, the target object stability is determined to be the first value. When the first indicator is greater than or equal to the first ratio threshold, or the second indicator is greater than or equal to the first ratio threshold, it indicates that the proportion of target objects with short life cycles is too high, and it may be that the radar performance has been greatly reduced or degraded, so the target object stability can be determined to be the first value; the first value indicates that the target object stability is extremely poor.

[0143] In some embodiments, when the first indicator is less than the first ratio threshold and greater than or equal to the second ratio threshold, and the second indicator is less than the first ratio threshold and greater than or equal to the second ratio threshold, the target object stability is determined to be the first value; wherein the second ratio threshold is less than the first ratio threshold. In other words, although the proportion of target objects with short life cycles represented by the first indicator does not reach the first ratio threshold, and the proportion of target objects with abnormal motion speeds does not reach the first ratio threshold, when both the first indicator and the second indicator are less than the first ratio threshold and greater than or equal to the second ratio threshold, it means that there are more target objects with short life cycles and abnormal motion speeds at the same time, which may be due to a significant reduction or degradation in radar performance, so the target object stability can still be determined to be the first value.

[0144] In some embodiments, when the first indicator is less than the first ratio threshold and greater than or equal to the second ratio threshold, or when the second indicator is less than the first ratio threshold and greater than or equal to the second ratio threshold, the target object stability is determined to be the second value. That is to say, when only one of the first indicator or the second indicator is less than the first ratio threshold and greater than or equal to the second ratio threshold, and the other indicator of the first indicator or the second indicator is normal, such as less than the second ratio threshold, it means that there is only one abnormal situation, such as a target object with a short life cycle, or a target object with abnormal movement speed. At this time, it may be that the radar performance has been reduced or degraded to a certain extent, so the target object stability can be determined to be the second value. The second value represents the poor stability of the target object. It can be seen that the stability of the target object represented by the second value is higher than the first value.

[0145] Exemplarily, the spliced ​​point cloud (point cloud data) is obtained from the synchronization message queue and sent to the pre-trained target detection network as input through the data container for real-time detection. The target detection network detects static and dynamic obstacles (target objects) on the output road in real time, as well as key attributes of static and dynamic obstacles such as center point coordinates (x, y, z), heading angle, object identification (id), life cycle (age), speed (lateral speed vx, longitudinal speed vy), etc. The target object stability assessment can be calculated and judged in the following way: First indicator: The proportion of target objects (first target objects) with a life cycle of <5 frames, and the statistical window is 10 seconds (first preset window). In some embodiments, the target number window can also be limited to 100 to save computing power. Second indicator: When the current speed, such as the vehicle speed ≥40km / h (first speed threshold), the target objects in the front (target motion direction) are accumulated in the second preset window, and the number of target objects (second target objects) with a speed of <1m / s (second speed threshold) is counted. In some embodiments, the maximum number of targets participating in the statistics can also be limited to 100 to save computing power. Taking the first ratio threshold as 0.8 and the second ratio threshold as 0.4 as an example, if the first index ≥ 0.8 and / or the second index ≥ 0.8, the output target object has extremely poor stability. If 0.4 ≤ the first index < 0.8, and 0.4 ≤ the second index < 0.8, the output target object has extremely poor stability. If 0.4 ≤ the first index < 0.8, or 0.4 ≤ index 2 < 0.8, the output target object has poor stability.

[0146] It can be understood that by detecting the stability of the target object in real time through point cloud data, the degradation of radar performance can be evaluated from the dimension of the radar scanning target object, so as to timely discover the degradation of radar performance so as to take targeted measures, thereby improving the real-time performance of radar detection and the stability of radar functions.

[0147] In some embodiments, the process of determining the missing point by performing ground detection on the point cloud data can be as follows: Figure 5 As shown, including:

[0148] S301, classify and detect the point cloud data through a ground detection network to determine the ground points in the point cloud data.

[0149] In some embodiments, the ground detection network may include a pre-trained classification network for distinguishing ground points from non-ground points in point cloud data. Exemplarily, the ground detection network may be implemented as a shallow binary classification network, trained with pre-labeled data, capable of outputting binary classification of input point cloud data, and labeling classification information for each point in the point cloud, such as a ground point or a non-ground point.

[0150] S302: Determine a target area in the field of view of the point cloud data, and determine the historical average number of sampling points in the target area.

[0151] In S302, the field of view (FOV) of the point cloud data corresponds to the coverage area of ​​the radar scanning range. For example, for the point cloud data obtained by splicing the original point cloud of the left radar and the original point cloud of the right radar, the field of view of the point cloud data may include the scanning range of the left radar and the scanning range of the right radar. The target area is determined in the field of view of the point cloud data, and the target area may include part or all of the area of ​​the preset position and size in the field of view. For the point cloud data obtained by the current scan, the number of sampling points of at least one historical radar scan in the target area can be counted and recorded in the previous radar scanning history process, so as to determine the historical average number of sampling points in the target area.

[0152] For example, the target area may include an area in the FOV of the point cloud data where Z=0 (ground height), 28≤x≤32m, and -2≤y≤2m in the Euclidean space. The target area corresponding to the forward-looking radar image may be projected as follows: Figure 6 shown.

[0153] S303, counting the number of echo points and the number of ground points in the target area of ​​the point cloud data, and determining the point loss situation based on the historical average number of sampling points, the number of echo points and the number of ground points.

[0154] In S303, the number of echo points in the target area of ​​the point cloud data obtained by the current scan and the number of ground points obtained by the ground detection network are counted, wherein the number of echo points may include the number of ground points and the number of non-ground points in the target area. For example, for a point cloud data, the data relationship diagram of the ground points, echo points and sampling points may be as follows: Figure 7 The missing points are determined based on the historical average number of sampling points, echo points, and ground points.

[0155] In some embodiments, the point loss situation is determined based on the historical average number of sampling points, the number of echo points, and the number of ground points, including:

[0156] When it is determined that the point cloud data meets the preset point loss condition based on the historical average number of sampling points, the number of echo points and the number of ground points, the point cloud data is recorded, and it is determined whether the next point cloud data meets the preset point loss condition; if the recorded point cloud data that meets the preset point loss condition is greater than or equal to the preset first number threshold, the ground point loss is determined to be a point loss state;

[0157] If the recorded point cloud data satisfying the preset point loss condition is less than or equal to the preset second numerical threshold, the current state of the ground point loss situation is determined to be a non-point loss state; the preset second numerical threshold is less than the preset first numerical threshold.

[0158] For example, the spliced ​​point cloud data is obtained from the synchronization message queue, and the pre-trained ground detection network is input through the data container to obtain the classified ground point P g and non-ground points; by clipping the field of view coverage of the spliced ​​radar point cloud, a ground point statistical range area is formed, that is, the target area G a By targeting the target area G a The actual historical point cloud data is counted to obtain the target area G a The historical average sampling points P sample When performing point loss detection, if Figure 8 As shown, the current point cloud data is obtained in the target area G a The echo point P in r , first determine the echo point P r The historical average sampling points P sample If the ratio This indicates that the number of echo points in the current point cloud data is significantly reduced compared to the historical average sampling points, and there may be missing points. At this time, it is determined that the current point cloud data meets the preset missing point condition and the point cloud data is added to the queue for subsequent multi-frame judgment. and This means that although the number of echo points has not decreased significantly compared to the historical average sampling points, the number of ground points is close to the total number of echo points, and non-ground points may be lost. In this case, they will also be added to the queue and wait for subsequent multi-frame judgment. and It indicates that there is no serious point loss in the current point cloud data, so the point cloud data is discarded and not recorded in the queue. Next, the point cloud data scanned by the radar is judged by multi-frame accumulation. For example, the above process is used to determine whether the accumulated 100 point cloud data obtained by the radar scan meet the preset point loss condition, and the point cloud data that meet the preset point loss condition is recorded in the queue. If the last point loss situation determined was no point loss, the preset first number threshold can be set to 80 (frames). When the number of point cloud data recorded in the queue that meet the preset point loss condition is greater than or equal to 80, the current point loss situation is determined to be point loss. If the last point loss situation determined was point loss, the preset second number threshold can be set to 20 (frames). When the radar has cumulatively scanned 100 point cloud data, the number of point cloud data recorded in the queue that meet the preset point loss condition is less than or equal to 20, and the current point loss situation is determined to be no point loss. Among them, the initial point loss situation can be pre-set to no point loss.

[0159] It can be understood that by evaluating the radar's point loss through ground point detection, the reduction or degradation of radar performance can be discovered in time for targeted processing, thereby improving the real-time performance of radar detection and the reliability of radar applications.

[0160] In some embodiments, the process of determining the noise situation by performing noise detection on the point cloud data may include:

[0161] S401-S402, such as Fig. 9 As shown below:

[0162] S401, performing noise detection on point cloud data through a noise detection network to determine the number of noise points in the point cloud data.

[0163] In some embodiments, the noise detection network can be a classification network trained using sample images pre-labeled with various types of noise, such as expansion noise, high-reflection noise, water splash dust exhaust noise, etc. By performing noise detection on point cloud data through the noise detection network, the confidence level of each point in the point cloud data belonging to a noise point can be obtained through classification prediction, thereby determining the number of noise points in the point cloud data.

[0164] S402: Determine the noise situation based on the number of noise points and the total number of points in the point cloud data.

[0165] In S402, based on the number of noise points and the total number of points in the point cloud data, the proportion of the number of noise points in the total number of points in the current point cloud data can be determined, and the proportion can reflect the noise situation of the point cloud data.

[0166] In some embodiments, the noise situation can be determined by accumulating multiple frames based on the number of points in the noise detection result and the number of points in the point cloud data, as follows:

[0167] Determine the noise condition of multiple frames of point cloud data containing point cloud data; for multiple frames of point cloud data, when the ratio between the number of noise points and the total number of points is greater than or equal to the noise point ratio threshold, and the number of point cloud data is greater than or equal to a preset third number threshold, determine that the noise condition is high noise.

[0168] Exemplarily, the point cloud data obtained after splicing is obtained from the synchronous message queue, and is sent to the pre-trained noise detection network through the data container input for real-time detection to obtain the noise detection result. The noise detection result includes the number of noise points in the current point cloud data. The noise detection result is added to the historical detection container. Exemplarily, the historical detection container can be a storage space that can accommodate 1,000 point cloud data. The noise detection result of each point cloud data added to the historical detection container is counted in real time, and the number of noise points pc in the noise detection result is counted. noise The total number of points in the point cloud data pc input If pc noise / pc input ≥0.2, (0.2 is the noise point ratio threshold), and if multiple (e.g. 100) point cloud data meet this condition, it is judged as high noise. Otherwise, it is judged as low noise.

[0169] It can be understood that by detecting the proportion of noise points in the point cloud data through the noise detection network, the reduction or degradation of radar performance can be discovered in time for targeted processing, thereby improving the real-time performance of radar detection and the reliability of radar applications.

[0170] In some embodiments, at least one environmental data may include: a wiper gear position signal on the vehicle; the above-mentioned process of performing rain detection based on the wiper gear position signal in the at least one environmental data and determining the rain situation may be as follows: Fig.10 As shown, including:

[0171] S501. When the wiper gear position signal indicates automatic gear, obtain at least one of a wiper speed and a wiper operation time.

[0172] In S501, a wiper gear position signal on the vehicle is obtained. If the wiper gear position signal indicates automatic gear, it means that the wiper of the current vehicle is operating automatically, and the radar may be currently operating in rain or snow. At least one of the wiper speed signal and the wiper operation time is obtained to determine the amount of rain and snow.

[0173] S502: Determine the rainfall according to at least one of the wiper speed and the wiper operation time.

[0174] In some embodiments, the rainfall condition may be determined based on the numerical ranges corresponding to different wiper speed signals; and / or the rainfall condition may be determined based on the wiper operation time. For example, the rainfall condition may indicate whether the rainfall is heavy rain, moderate rain, or light rain.

[0175] Exemplarily, the wiper information bcm is obtained from the synchronous message queue msg ; Wiper information bcm msg It may include a wiper gear signal and a wiper speed. sts Can include any of the 13 preset wiper gears. msg The rainfall situation judgment module is input through the data container. First, based on the wiper gear signal wiper sts If the wiper gear position signal indicates automatic gear, for example, wiper sts = 0x8 is automatic gear configuration, and the wiper speed is obtained from bcm_msg speed Signal, if 0<wiper speed <3, and the wiper operation time exceeds 30s, the rainfall is determined to be light rain. speed <6, and the wiper operation time exceeds 30s, the rainfall is determined to be moderate rain. speed ≥6, and the wiper operation time exceeds 30s, the rainfall is determined to be heavy rain.

[0176] In some embodiments, the above-mentioned process of detecting the amount of rain based on the wiper gear position signal in at least one environmental data and determining the amount of rain can also be as follows: Fig.11 As shown, including:

[0177] S601: When the wiper gear position signal indicates a manual gear, obtain a manual gear trigger signal and a wiper speed.

[0178] In S601, when the wiper gear position signal indicates manual gear, it means that the user is manually operating the wiper. A manual gear trigger signal and a wiper speed are obtained.

[0179] S602: Determine the wiper triggering times according to the manual gear triggering signal.

[0180] In S602, each time the wiper is manually operated, a manual gear trigger signal is generated. The number of wiper trigger times can be determined according to the manual gear trigger signal to represent the operating frequency of the wiper.

[0181] S603: Determine the rainfall according to the number of wiper triggering times.

[0182] In S603, the number of times the windshield wiper is triggered can reflect the frequency of the user's manual operation of the windshield wiper. Therefore, based on the number of times the windshield wiper is triggered, the rainfall situation in the current environment can be determined.

[0183] Exemplarily, obtain the windshield wiper information bcm from the synchronization message queue msg ; The windshield wiper information bcm msg may include the windshield wiper gear signal wiper sts and the windshield wiper speed wiper speed . Exemplarily, the windshield wiper gear signal wiper sts may include any one of 13 preset windshield wiper gears. Create a windshield wiper signal container wiper_sts_history[k] for storing the wiper synchronized with the acquired historical point cloud data sts and the wiper synchronized with the current point cloud data sts . Where k is an integer not less than 0. Input the windshield wiper information bcm msg into the rain and snow amount situation judgment module through the data container. First, based on the windshield wiper gear signal wiper sts judge. If wiper_sts = 0xB, it indicates that the windshield wiper is in the off state or the manual gear. At this time, if it is judged that the windshield wiper is in the stationary state and the duration exceeds 30s, it is determined that the rainfall situation is no rain. If a manual trigger signal for the user to manually trigger the windshield wiper is received, judge wiper_sts_history[k]: If wiper_sts_history[0] is 0xB, and 0 < wiper_sts_history[1] ≤ 7, and wiper_sts_history[2] = 0, and wiper_sts_history[3] = 0xB, it means that the user has manually triggered the operation of the windshield wiper once, and the count of the number of times the windshield wiper is triggered manual_moved_count is increased. Combine the windshield wiper speed wiper speed to judge. If manual_moved_count_ > 10, it is determined that the rainfall situation is light rain; if manual_moved_count > 500. And 0 < wiper speed < 3, it is determined that the rainfall situation is moderate rain. If 3 ≤ wiper speed < 6, it is determined that the rainfall situation is moderate rain. If wiper_speed ≥ 6, it is determined that the rainfall situation is heavy rain.

[0184] It can be understood that by detecting the rainfall situation in the radar working environment through the windshield wiper gear signal, the reduction or degradation of the radar performance can be discovered in time for targeted processing, thereby improving the real-time performance of radar performance detection and the reliability of radar applications.

[0185] In some embodiments, at least one environmental data includes: a radar window cleanliness signal. The radar window cleanliness signal includes: at least one partition on the radar window corresponds to at least one regional cleanliness signal; the above-mentioned process of detecting the obstruction degree of the radar window based on the radar window cleanliness signal in the at least one environmental data and determining the obstruction condition of the radar window can be as follows: Fig.12 As shown, including:

[0186] S701. Count regional cleanliness signals within at least one preset regional range on a radar window according to at least one regional cleanliness signal, and determine at least one regional statistical result corresponding to the at least one preset regional range.

[0187] In S701, the preset area range includes one or more partitions; for each preset area range, one or more area cleanliness signals corresponding to one or more partitions within the preset area range are counted to obtain the area statistical results corresponding to the preset area range, thereby determining at least one area statistical result corresponding to at least one preset area range.

[0188] In some embodiments, the radar window comprises a radome.

[0189] S702: Determine radar window obstruction status according to at least one regional statistical result.

[0190] In S702, based on the statistical result of at least one region, it can be determined whether each region on the radar window is blocked by dirt or foreign objects, that is, the radar window blocking condition is determined.

[0191] For example, the laser radar hood surface is divided into 8x12 regions according to the laser radar horizontal FOV angle interval and the vertical FOV angle interval. The regional cleanliness signal feature flag of each region in the 8x12 matrix region is obtained from the synchronization message queue to equivalently characterize the severity of the same type of dirt; the feature flag of the 8x12 matrix region may include f 1,1 to f 12,8 ,as follows:

[0192]

[0193] The contamination status of the 8 channel intervals in the horizontal interval is marked with 2 bytes, which represent the channel interval ID (0 to 8) from low to high. The contamination status of each channel interval is represented by 2 bits, and bit [0] is the transparent contamination diagnosis fault status D trans , 1 indicates that penetrating dirt occurs, 0 indicates that there is no penetrating dirt, bit [1] non-penetrating dirt diagnosis fault status Dnon-trans , 1 means that non-transparent dirt has occurred, 0 means that non-transparent dirt has not occurred, and other horizontal intervals ID1 to 11 are consistent with ID0; the feature flag of the 8x12 area is parsed, and if any interval state is not equal to 0, it is determined that there is attachment on the surface of the hood, and the feature flag in the specified area class is counted. The statistical count mainly distinguishes between non-ROI areas and ROI areas. The ROI area is the vertical ID (4, 5, 6, 7, 8, 9) and the horizontal ID is (3, 4, 5, 6), and the others are non-ROI areas. When the ROI area feature flag = D trans , and the count value Count Dtrans / Count total ≥0.75, the radar window is judged to be severely dirty; 〖0.5≤Count〗_(D_trans) / 〖Count〗_total<0.75, the radar window is judged to be moderately dirty; when 0.5≤Count Dtrans / Count total <0.75, the radar window is judged to be slightly dirty.

[0194] It can be understood that by detecting the rainfall conditions in the radar working environment through the wiper signal, the reduction or degradation of radar performance can be discovered in time for targeted processing, thereby improving the real-time performance of radar detection and the reliability of radar application.

[0195] In some embodiments, the process of determining the weather scene by performing weather scene detection on the point cloud data can be as follows: Fig.13 As shown, including:

[0196] S801. Predict the point cloud data through a weather scene classification network to determine the classification result corresponding to the point cloud data.

[0197] In S801, an initial classification neural network can be trained by pre-labeled weather type data, such as labeled data of rainy days and non-rainy days, to obtain a scene classification network. The point cloud data is predicted by the weather scene classification network to obtain a day classification result corresponding to the point cloud data. The classification result represents the weather scene type corresponding to the point cloud data, such as whether the point cloud data belongs to a rainy day or a non-rainy day.

[0198] S802: Determine a weather scene by counting multiple classification results corresponding to multiple point cloud data.

[0199] In S802, multiple classification results of multiple point cloud data are accumulated, and classification statistics are performed on the multiple classification results, so as to determine a weather scene greater than a preset scene classification threshold as the weather scene of the radar working environment.

[0200] Exemplarily, the point cloud data obtained after splicing is obtained from the synchronization message queue, and is sent to the pre-trained scene classification network through the data container input for real-time detection, and the classification result scene_result of the point cloud data is output, and the classification result scene_result of the point cloud data is added to the classification result container, and the classification result container is used to store multiple classification results corresponding to multiple point cloud data. The classification results in the classification result container are traversed and counted to obtain the proportion of rainy day results in the multiple classification results. If the proportion of rainy day results is ≥0.8 (0.8 is the preset scene classification threshold), it is judged as a rainy day, otherwise it is defined as non-raining.

[0201] It can be understood that by detecting the weather scene type of the radar working environment through the weather scene classification network signal, the reduction or degradation of radar performance can be discovered in time for targeted processing, thereby improving the real-time performance detection of radar and the reliability of radar application.

[0202] In some embodiments, based on the above-mentioned target object stability detection, point loss detection, weather scene detection, noise detection, rainfall detection, radar window occlusion degree detection and other processes, such as Fig.14 As shown, by time synchronization and preprocessing the input laser radar window cleanliness signal, vehicle chassis signal (mainly including vehicle speed, equivalent to the current speed in the aforementioned embodiment), left radar original point cloud, right radar original point cloud and wiper sensor signal, point cloud data and at least one environmental data are obtained. According to the point cloud data and at least one environmental data, the radar performance detection method provided in the embodiment of the present application is used to perform multi-channel processing of target object detection, wiper signal judgment, laser radar hood cleanliness signal statistics, ground point detection, noise segmentation network and scene classification network classification weather scene to obtain a multi-channel processing result. Among them, the multi-channel processing results include: the stability of the target object obtained through target object detection; the classification of rainfall determined by the wiper signal (equivalent to the rainfall situation); the radar window occlusion obtained by the statistics of the laser radar hood cleanliness signal; the ground point loss rate statistics of the current point cloud data through ground point detection, and the point loss situation determined by accumulating the ground point loss rate of multiple point cloud data; noise segmentation through the noise segmentation network, and the noise situation obtained by the number of noise points obtained by statistical segmentation, and the classification of weather scenes obtained by the scene classification network; finally, the current degradation of the radar is determined by integrating the multi-channel processing results (equivalent to the detection based on the point cloud data and at least one environmental data, determining the stability of the target object scanned by the radar, the point loss situation and at least one of the environmental detection results). Exemplarily, according to the combination of different processing results, flag bits of different degrees of degradation can be output to characterize the degradation situation, including:

[0203] R1: indicates heavy rainfall;

[0204] R2: indicates moderate rainfall and ground points are lost;

[0205] R3: indicates that the weather scene is classified as rainy day with moderate rainfall;

[0206] R4: indicates that the rainfall is moderate and the amount of noise is large;

[0207] R5: indicates that the target object has extremely poor stability;

[0208] R6: indicates that the weather scene is classified as rainy and the ground point is missing;

[0209] R7: indicates that the weather scene is classified as rainy day and the target object has poor stability;

[0210] R8: indicates moderate rainfall and much obstruction of the radar window;

[0211] RX: indicates other rules.

[0212] Exemplarily, RX may include:

[0213] R9: indicates that the stability of the target object is poor and the rainfall is moderate;

[0214] R10: Indicates that the target object has poor stability and the radar window is heavily obstructed;

[0215] R11: indicates that the target object has poor stability and ground points are lost;

[0216] R12: indicates that the target object has poor stability and a large amount of noise;

[0217] R13: Indicates that the radar window is too blocked and the ground point is lost;

[0218] R14: The radar window is too blocked and the scene is classified as rainy day

[0219] R15: The radar window is too blocked and has many noise points;

[0220] R16: ground points are missing and the number of noises is large;

[0221] R17: The weather scene is classified as rainy day, and there are many noise points.

[0222] More other rules can be combined, which are not listed here. If the predefined degradation rule is not met, a result indicating that the radar performance is not degraded is output.

[0223] It can be understood that the embodiment of the present invention accesses the original point cloud information and the vehicle's own windshield wiper rain sensor, the window occlusion flag of the laser radar and other signals, and realizes the judgment of the performance degradation of the laser radar algorithm through the combination of traditional and deep learning algorithms, and outputs a judgment signal to the back end. In this way, when the radar performance degradation is detected in time, the back-end module can use and process the signal output by the laser radar algorithm in a targeted manner according to the performance degradation signal provided by this method, so as to ensure the robustness and safety of the system.

[0224] The embodiment of the present invention provides a radar performance detection device 1, such as Fig.15 As shown, including:

[0225] An acquisition module 11, used to determine the point cloud data scanned by the radar, and to acquire at least one environmental data synchronized with the point cloud data;

[0226] A detection module 12, configured to perform detection based on the point cloud data and / or the at least one environmental data to determine at least one of the stability of the target object scanned by the radar, the point loss situation, and the environmental detection result;

[0227] The determination module 13 is used to determine the degradation of radar performance according to at least one of the target object stability, the point loss situation and the environmental detection result.

[0228] In some embodiments, the acquisition module 11 is also used to acquire at least one original point cloud scanned by each radar among multiple radars; determine multiple synchronized original point clouds corresponding to the multiple radars based on at least one time information corresponding to the at least one original point cloud; and splice the multiple synchronized original point clouds to determine the point cloud data.

[0229] In some embodiments, the acquisition module 11 is also used to acquire at least one sensor data stream collected by at least one sensor; for each sensor data stream of the at least one sensor data stream, the sensor data synchronized with the point cloud data in each sensor data stream is determined according to the time information of the point cloud data, as the environmental data synchronized with the point cloud data, thereby determining the at least one environmental data.

[0230] In some embodiments, the environmental detection results include: at least one of weather scenes, noise conditions, rainfall conditions and radar window obstruction conditions; the detection module 12 is also used to perform target object detection based on the point cloud data to determine the stability of the target object; and / or, perform ground detection based on the point cloud data to determine the point loss situation; and / or, determine the weather scene by performing weather scene detection on the point cloud data; and / or, determine the noise condition by performing noise detection on the point cloud data; and / or, perform rainfall detection based on the wiper gear signal in the at least one environmental data to determine the rainfall condition; and / or, perform radar window obstruction degree detection based on the radar window cleanliness signal in the at least one environmental data to determine the radar window obstruction condition.

[0231] In some embodiments, the detection module 12 is also used to perform target detection on the point cloud data in the first point cloud data set, determine the total number of target objects corresponding to the first point cloud data set, and the first number of first target objects whose life cycle is less than or equal to a preset life cycle threshold corresponding to the first point cloud data set; the first point cloud data set includes at least one point cloud data within a first time window; based on the first number and the total number of target objects corresponding to the first point cloud data set, determine a first indicator; and / or, based on at least one environmental data, determine a current speed; when the current speed is greater than or equal to a first speed threshold, determine the total number of target objects corresponding to the second point cloud data set and the second number of second target objects whose speed in the target motion direction is less than a second speed threshold corresponding to the second point cloud data set by performing target detection on the point cloud data in the second point cloud data set; the second point cloud data set includes at least one point cloud data within a second time window; based on the second number and the total number of target objects corresponding to the second point cloud data set, determine a second indicator; based on at least one of the first indicator and the second indicator, determine the stability of the target object.

[0232] In some embodiments, the first indicator represents the ratio of the first number to the total number of target objects corresponding to the first point cloud data set; the second indicator represents the ratio of the second number to the total number of target objects corresponding to the second point cloud data set, and the detection module 12 is also used to determine that the stability of the target object is a first value when at least one of the first indicator and the second indicator is greater than or equal to a first ratio threshold; the first value represents that the stability of the target object is extremely poor; when the first indicator is less than the first ratio threshold and greater than or equal to the second ratio threshold, and the second indicator is less than the first ratio threshold and greater than or equal to the second ratio threshold, the stability of the target object is determined to be the first value; the second ratio threshold is less than the first ratio threshold; when the first indicator is less than the first ratio threshold and greater than or equal to the second ratio threshold, or the second indicator is less than the first ratio threshold and greater than or equal to the second ratio threshold, the stability of the target object is determined to be a second value; the second value represents that the stability of the target object is poor.

[0233] In some embodiments, the detection module 12 is also used to classify and detect the point cloud data through a ground detection network to determine the ground points in the point cloud data; determine the target area in the field of view of the point cloud data, and determine the historical average number of sampling points in the target area; count the number of echo points and the number of ground points of the point cloud data in the target area, and determine the point loss situation based on the historical average number of sampling points, the number of echo points and the number of ground points.

[0234] In some embodiments, the detection module 12 is also used to record the point cloud data and determine whether the next point cloud data meets the preset point loss condition when it is determined that the point cloud data meets the preset point loss condition based on the historical average number of sampling points, the number of echo points and the number of ground points; if the recorded point cloud data that meets the preset point loss condition is greater than or equal to the preset first number threshold, the ground point loss situation is determined to be a point loss state; if the recorded point cloud data that meets the preset point loss condition is less than or equal to the preset second number threshold, the current state of the ground point loss situation is determined to be a non-point loss state; the preset second number threshold is less than the preset first number threshold.

[0235] In some embodiments, the detection module 12 is also used to perform noise detection on the point cloud data through a noise detection network to determine the number of noise points in the point cloud data; and determine the noise situation based on the number of noise points and the total number of points in the point cloud data.

[0236] In some embodiments, the detection module 12 is also used to determine the noise condition of multiple frames of point cloud data containing the point cloud data; for the multiple frames of point cloud data, when the ratio between the number of noise points and the total number of points is greater than or equal to the noise point ratio threshold, and the number of point cloud data is greater than or equal to a preset third number threshold, the noise condition is determined to be high noise.

[0237] In some embodiments, the detection module 12 is further used to obtain at least one of a wiper speed and a wiper operating time when the wiper gear position signal indicates an automatic gear; and determine the rainfall condition based on at least one of the wiper speed and the wiper operating time.

[0238] In some embodiments, the detection module 12 is further used to obtain a manual gear trigger signal and a wiper speed when the wiper gear position signal indicates a manual gear; determine the number of wiper trigger times according to the manual gear trigger signal; and determine the rainfall conditions according to the number of wiper trigger times.

[0239] In some embodiments, the cleanliness signal includes: at least one partition on the radar window corresponds to at least one regional cleanliness signal; the detection module 12 is also used to count the regional cleanliness signals within at least one preset regional range on the radar window according to the at least one regional cleanliness signal, and determine at least one regional statistical result corresponding to the at least one preset regional range; the preset regional range includes one or more partitions; and the radar window obstruction condition is determined according to the at least one regional statistical result.

[0240] In some embodiments, the detection module 12 is further used to predict the point cloud data through a weather scene classification network to determine a classification result corresponding to the point cloud data; and to determine the weather scene by counting multiple classification results corresponding to multiple point cloud data.

[0241] It should be noted that the description of the above device embodiment is similar to the description of the above method embodiment, and has similar beneficial effects as the method embodiment. For technical details not disclosed in the device embodiment of the present invention, please refer to the description of the method embodiment of the present invention for understanding.

[0242] An embodiment of the present invention further provides an electronic device, Fig.16 This is an optional structural diagram of an electronic device 3 provided in an embodiment of the present invention. Fig.16 As shown, the electronic device 3 includes: a memory 32 and a processor 33. The memory 32 and the processor 33 are connected via a communication bus 34; the memory 32 is used to store executable instructions; the processor 33 is used to implement the radar performance detection method provided by the embodiment of the present invention when executing the executable instructions stored in the memory 32.

[0243] An embodiment of the present invention provides a vehicle, the vehicle comprising: Fig.15 The radar performance detection device shown, or, including Fig.16 Electronic equipment shown.

[0244] An embodiment of the present invention provides a computer-readable storage medium storing executable instructions, wherein executable instructions are stored. When the executable instructions are executed by the above-mentioned processor, the above-mentioned processor will be caused to execute the radar performance detection method provided by the embodiment of the present invention.

[0245] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface storage, optical disk, or CD-ROM; or it may be various devices including one or any combination of the above memories.

[0246] In some embodiments, executable instructions may be in the form of a program, software, software module, script or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a stand-alone program or as a module, component, subroutine or other unit suitable for use in a computing environment.

[0247] As examples, executable instructions may, but need not, correspond to a file in a file system, may be stored as part of a file storing other programs or data, such as one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files storing one or more modules, subroutines, or code portions). As examples, executable instructions may be deployed to be executed on one computing device, or on multiple computing devices located at one location, or on multiple computing devices distributed at multiple locations and interconnected by a communication network.

[0248] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0249] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0250] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0251] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0252] The above embodiments are only preferred embodiments for fully illustrating the present invention, and the protection scope of the present invention is not limited thereto. Any equivalent substitution or change made by a person skilled in the art based on the present invention is within the protection scope of the present invention.

Claims

1. A radar performance detection method, characterized in that: include: Determine point cloud data scanned by the radar, and obtain at least one environmental data synchronized with the point cloud data; Perform detection based on the point cloud data and / or the at least one environmental data to determine at least one of the stability of the target object scanned by the radar, the point loss situation, and the environmental detection result; Determine the radar performance degradation condition according to at least one of the target object stability, the point loss condition, and the environment detection result.

2. The method according to claim 1, characterized in that: The step of determining the point cloud data scanned by the radar comprises: Acquire at least one original point cloud scanned by each radar of the plurality of radars; Determining a plurality of synchronized original point clouds corresponding to the plurality of radars according to at least one time information corresponding to the at least one original point cloud; The multiple synchronized original point clouds are spliced ​​to determine the point cloud data.

3. The method according to claim 1, characterized in that The acquiring of at least one environmental data synchronized with the point cloud data comprises: Obtaining at least one sensor data stream collected by at least one sensor; For each sensor data stream of at least one sensor data stream, sensor data synchronized with the point cloud data in each sensor data stream is determined according to time information of the point cloud data as environmental data synchronized with the point cloud data, thereby determining the at least one environmental data.

4. The method according to any one of claims 1 to 3, characterized in that: The environmental detection result includes: at least one of weather scene, noise condition, rainfall condition and radar window occlusion condition; the radar working environment detection and / or radar scanning effect detection is performed based on the point cloud data and / or the at least one environmental data to determine at least one of the stability of the target object scanned by the radar, the point loss condition and the environmental detection result, including: Performing target object detection based on the point cloud data to determine the stability of the target object; And / or, performing ground detection based on the point cloud data to determine the point loss situation; and / or, determining the weather scene by performing weather scene detection on the point cloud data; And / or, determining the noise condition by performing noise detection on the point cloud data; and / or, performing rain detection based on a wiper gear position signal in the at least one environmental data to determine the rain condition; And / or, based on the radar window cleanliness signal in the at least one environmental data, the radar window obstruction degree detection is performed to determine the radar window obstruction condition.

5. The method according to claim 4, characterized in that The step of determining the stability of the target object by performing target object detection on the point cloud data includes: Performing target detection on the point cloud data in the first point cloud data set, determining the total number of target objects corresponding to the first point cloud data set, and the first number of first target objects whose life cycles are less than or equal to a preset life cycle threshold value corresponding to the first point cloud data set; the first point cloud data set includes at least one point cloud data within a first time window; determining a first indicator based on the first number and the total number of target objects corresponding to the first point cloud data set; and / or, determining a current speed based on at least one environmental data; In a case where the current speed is greater than or equal to the first speed threshold, determining the total number of target objects corresponding to the second point cloud data set and the second number of second target objects in the target motion direction corresponding to the second point cloud data set and having a speed less than the second speed threshold by performing target detection on the point cloud data in the second point cloud data set; the second point cloud data set includes at least one point cloud data within the second time window; determining a second indicator based on the second number and the total number of target objects corresponding to the second point cloud data set; The target object stability is determined based on at least one of the first indicator and the second indicator.

6. The method according to claim 5, characterized in that The first indicator represents a ratio of the first number to the total number of target objects corresponding to the first point cloud data set; The second indicator represents a ratio of the second number to the total number of target objects corresponding to the second point cloud data set, and the determining the stability of the target object based on at least one of the first indicator and the second indicator includes: When at least one of the first indicator and the second indicator is greater than or equal to a first ratio threshold, determining that the target object stability is a first value; the first value indicates that the target object stability is extremely poor; In the case where the first indicator is less than the first ratio threshold and greater than or equal to the second ratio threshold, and the second indicator is less than the first ratio threshold and greater than or equal to the second ratio threshold, determining that the target object stability is a first value; and the second ratio threshold is less than the first ratio threshold; When the first indicator is less than the first ratio threshold and greater than or equal to the second ratio threshold, or when the second indicator is less than the first ratio threshold and greater than or equal to the second ratio threshold, the target object stability is determined to be a second value; the second value indicates that the target object has poor stability.

7. The method according to claim 4, characterized in that The step of determining the point loss situation by performing ground detection on the point cloud data includes: Classify and detect the point cloud data through a ground detection network to determine ground points in the point cloud data; Determine a target area in the field of view of the point cloud data, and determine a historical average number of sampling points in the target area; The number of echo points and the number of ground points of the point cloud data in the target area are counted, and the point loss situation is determined according to the historical average number of sampling points, the number of echo points and the number of ground points.

8. The method according to claim 7, characterized in that The determining the point loss situation according to the historical average sampling point number, the echo point number and the ground point number includes: When it is determined that the point cloud data meets the preset point loss condition according to the historical average sampling point number, the echo point number and the ground point number, the point cloud data is recorded, and it is determined whether the next point cloud data meets the preset point loss condition; If the recorded point cloud data satisfying the preset point loss condition is greater than or equal to a preset first quantity threshold, then determining that the ground point loss situation is a point loss state; If the recorded point cloud data that meets the preset point loss condition is less than or equal to the preset second quantity threshold, the current state of the ground point loss situation is determined to be a no-point loss state; the preset second quantity threshold is less than the preset first quantity threshold.

9. The method according to claim 4, characterized in that The step of performing noise detection on the point cloud data to determine the noise situation includes: Performing noise detection on the point cloud data through a noise detection network to determine the number of noise points in the point cloud data; The noise situation is determined based on the number of noise points and the total number of points in the point cloud data.

10. The method according to claim 9, characterized in that The determining the noise situation based on the number of noise points and the total number of points in the point cloud data includes: Determine the noise condition of multiple frames of point cloud data including the point cloud data; For the multi-frame point cloud data, when the ratio between the number of noise points and the total number of points is greater than or equal to the noise point ratio threshold, and the number of point cloud data is greater than or equal to a preset third number threshold, the noise situation is determined to be high noise.

11. The method according to claim 4, characterized in that The performing rain detection based on the wiper gear position signal in the at least one environmental data to determine the rain situation includes: When the wiper gear position signal indicates automatic gear, obtaining at least one of a wiper speed and a wiper operation time; The rainfall condition is determined according to at least one of the wiper speed and the wiper operation time.

12. The method according to claim 4, characterized in that The performing rain detection based on the wiper gear position signal in the at least one environmental data to determine the rain situation includes: When the wiper gear position signal indicates a manual gear, obtaining a manual gear trigger signal and a wiper speed; Determining the wiper triggering times according to the manual gear triggering signal; The rainfall condition is determined according to the wiper triggering times.

13. The method according to claim 4, characterized in that The cleanliness signal includes: at least one partition on the radar window corresponds to at least one regional cleanliness signal; the radar window obstruction degree detection based on the radar window cleanliness signal in the at least one environmental data to determine the radar window obstruction condition includes: According to the at least one regional cleanliness signal, statistically analyze the regional cleanliness signals within at least one preset regional range on the radar window, and determine at least one regional statistical result corresponding to the at least one preset regional range; the preset regional range includes one or more partitions; The radar window obstruction condition is determined according to the at least one regional statistical result.

14. The method according to claim 4, characterized in that The determining the weather scene by performing weather scene detection on the point cloud data includes: Predicting the point cloud data through a weather scene classification network to determine a classification result corresponding to the point cloud data; The weather scene is determined by counting multiple classification results corresponding to multiple point cloud data.

15. A radar performance detection device, characterized in that: The device comprises: An acquisition module, used to determine point cloud data scanned by the radar, and to acquire at least one environmental data synchronized with the point cloud data; A detection module, configured to perform detection based on the point cloud data and / or the at least one environmental data to determine at least one of stability of a target object scanned by the radar, a point loss situation, and an environmental detection result; A determination module is used to determine the degradation of radar performance according to at least one of the target object stability, the point loss situation and the environmental detection result.

16. An electronic device, characterized in that: The electronic device comprises: A memory for storing computer executable instructions; A processor, configured to implement the method according to any one of claims 1 to 14 when executing the computer executable instructions stored in the memory.

17. A computer-readable storage medium storing computer-executable instructions, characterized in that: When the computer executable instructions or computer program are executed by a processor, the method according to any one of claims 1 to 14 is implemented.

18. A computer program product comprising computer executable instructions, characterized in that: When the computer executable instructions or computer program are executed by a processor, the method according to any one of claims 1 to 14 is implemented.

19. A vehicle, characterized in that: The vehicle comprises: the radar performance detection device as claimed in claim 15, or the electronic device as claimed in claim 16.

Images