Processing method and device for target detection program of unmanned vehicle

Abstract

The present disclosure provides a method and device for processing an object detection program of an unmanned vehicle, relating to the field of unmanned driving technology. The method comprises: obtaining perception data of the unmanned vehicle, the perception data comprising multiple frames of point cloud data of multiple different scenes; obtaining target object true value calibration information of each frame of point cloud data of each scene, and target object detection results of the target object detection program for the frame of point cloud data; matching the target object detection results of the frame of point cloud data with the target object true value calibration information of the frame of point cloud data, determining the target object missed detection rate, target object false detection rate, and target object detection error of the frame of point cloud data based on the matching results, so as to determine the test sub-results of the frame of point cloud data; obtaining the target test results of each scene based on the test sub-results of the multiple frames of point cloud data of each scene; and determining a processing strategy for the target object detection program based on the target test results of multiple different scenes to process the detection program.

Description

Technical Field

[0001] The present disclosure relates to the field of unmanned driving technology, and in particular to a method and device for processing a target object detection program of an unmanned vehicle. Background Art

[0002] As a key component of smart mines, autonomous driving technology in mining areas improves production efficiency and promotes the safe and green development of the mining industry. Mining transport vehicles are equipped with sensing devices. The point cloud data sensed by the sensing devices is sent to a specific object detection program for point cloud recognition. Based on the recognition results, information about obstacles ahead can be determined, enabling effective driving control.

[0003] In the mining industry, the lack of structured semantic information such as lane markings, traffic lights, and traffic signs in the operating area places higher demands on object detection programs. However, the relevant technologies have not yet proposed a method for effectively verifying the correctness of such programs and conducting subsequent processing for this purpose. Summary of the Invention

[0004] The present disclosure provides a method and apparatus for processing an object detection program for an unmanned vehicle, which solves the problems in the related art. The technical solution is as follows:

[0005] In a first aspect, a method for processing an object detection program of an unmanned vehicle is provided, the method comprising:

[0006] Acquire perception data collected by the perception device of the unmanned vehicle, wherein the perception data includes multiple frames of point cloud data corresponding to multiple different scenes;

[0007] For each frame of point cloud data corresponding to each scene, perform the following operations:

[0008] Obtain the target object true value calibration information of the frame point cloud data;

[0009] Obtain the target object detection result of the target object detection program for the frame point cloud data;

[0010] Matching the target object detection result of the frame point cloud data with the target object true value calibration information of the frame point cloud data to obtain a matching processing result;

[0011] Determine, based on the matching processing result, a target object missed detection rate, a target object false detection rate, and a target object detection error corresponding to the frame point cloud data, and determine a test sub-result corresponding to the frame point cloud data based on the target object missed detection rate, the target object false detection rate, and the target object detection error;

[0012] Obtain the target test result corresponding to each scene based on the test sub-results of the multi-frame point cloud data corresponding to each scene;

[0013] A processing strategy for the target object detection program is determined based on the target test results corresponding to the multiple different scenarios, and the detection program is processed according to the processing strategy.

[0014] In a possible implementation, determining a test sub-result corresponding to the frame of point cloud data according to the target object missed detection rate, the target object false detection rate, and the target object detection error includes:

[0015] If the target object missed detection rate is less than a preset first threshold, the target object false detection rate is less than a preset second threshold, and the target object detection error is less than a preset third threshold, determining the test sub-result as passing the test on the perception capability of the target object detection program for the frame of point cloud data;

[0016] When the target object missed detection rate is greater than or equal to the preset first threshold, the target object false detection rate is greater than or equal to the preset second threshold, and / or the target object detection error is greater than or equal to the preset third threshold, the test sub-result is determined to be that the target object detection program's perception ability of the frame point cloud data has failed the test.

[0017] In a possible implementation, obtaining a target test result corresponding to each scene based on the test sub-results of the multi-frame point cloud data corresponding to each scene includes:

[0018] According to the test sub-results, the test pass rate of the point cloud data corresponding to the scene is obtained as the target test result corresponding to the scene.

[0019] In one possible implementation, determining a processing strategy for the target object detection program according to target test results corresponding to the multiple different scenarios includes:

[0020] Obtaining correction coefficients corresponding to the multiple different scenarios, where the magnitude of the correction coefficients is related to the degree of danger corresponding to the scenario;

[0021] Calculating a final test pass rate according to the test pass rates corresponding to the multiple different scenarios and the correction coefficients corresponding to the multiple different scenarios;

[0022] A processing strategy for the target object detection program is determined based on the final test pass rate.

[0023] In a possible implementation, matching the target object detection result of the frame point cloud data with the target object true value calibration information of the frame point cloud data to obtain a matching result includes:

[0024] An intersection-over-union ratio of the target object detection result of the frame point cloud data and the target object true value calibration information of the frame point cloud data is determined as the matching processing result.

[0025] In a possible implementation, the target object detection error corresponding to the frame point cloud data is determined according to the matching processing result in the following manner:

[0026] Determining true value calibration information of the target object that successfully matched in the frame of point cloud data based on the intersection-over-union ratio of the frame of point cloud data and a preset fourth threshold;

[0027] The distance difference between the true value calibration information of the target object that is successfully matched in the frame point cloud data and the corresponding target object detection result is calculated to obtain the target object detection error corresponding to the frame point cloud data.

[0028] In a possible implementation, the target object missed detection rate corresponding to the frame point cloud data is determined according to the matching processing result in the following manner:

[0029] Determining true value calibration information of the target object that was not successfully matched in the frame of point cloud data based on the intersection-over-union ratio of the frame of point cloud data and a preset fourth threshold;

[0030] Determine the number of target object true value calibration information that has not been successfully matched in the frame of point cloud data and the total number of target object true value calibration information in the frame of point cloud data;

[0031] The ratio between the number of the target object true value calibration information that was not successfully matched and the total number of the target object true value calibration information is used as the target object missed detection rate.

[0032] In one possible implementation, the target object false detection rate corresponding to the frame point cloud data is determined according to the matching processing result in the following manner:

[0033] Determining a detection result of an unsuccessful target object in the frame of point cloud data based on an intersection-over-union ratio of the frame of point cloud data and a preset fourth threshold;

[0034] Determine the number of unmatched target object detection results in the frame of point cloud data and the total number of target object detection results in the frame of point cloud data;

[0035] The ratio between the number of the unmatched target object detection results and the total number of the target object detection results is used as the target object false detection rate.

[0036] In a possible implementation, the multiple different scenarios include test scenarios corresponding to at least two different types of operating areas in a mining area; and / or,

[0037] The multiple different scenarios include test scenarios corresponding to at least two obstacle types.

[0038] In a second aspect, a processing device for an object detection program of an unmanned vehicle is provided, the device comprising:

[0039] The first acquisition unit is configured to acquire perception data collected by the perception device of the unmanned vehicle, wherein the perception data includes multiple frames of point cloud data corresponding to multiple different scenes. For each frame of point cloud data corresponding to each scene, the following units perform the following corresponding operations:

[0040] A second acquisition unit is used to obtain the target object true value calibration information of the frame point cloud data;

[0041] A third acquisition unit is used to obtain the target object detection result of the target object detection program on the frame point cloud data;

[0042] a first matching unit, configured to match the target object detection result of the frame point cloud data with the target object true value calibration information of the frame point cloud data to obtain a matching result;

[0043] a first determining unit, configured to determine a target object missed detection rate, a target object false detection rate, and a target object detection error corresponding to the frame point cloud data based on the matching processing result, and determine a test sub-result corresponding to the frame point cloud data based on the target object missed detection rate, the target object false detection rate, and the target object detection error;

[0044] A first obtaining unit is configured to obtain a target test result corresponding to each scene according to the test sub-results of the multi-frame point cloud data corresponding to each scene;

[0045] The second determining unit is configured to determine a processing strategy for the target object detection program according to the target test results corresponding to the multiple different scenarios, and process the detection program according to the processing strategy.

[0046] According to a third aspect, a computer-readable storage medium is provided, wherein the storage medium stores at least one instruction, and the at least one instruction is loaded and executed by a processor to implement the method according to the aspect described above.

[0047] In a fourth aspect, an electronic device is provided, including:

[0048] a memory for storing computer-readable instructions; and

[0049] A processor is configured to execute the computer-readable instructions so that the electronic device executes the method described above.

[0050] The beneficial effects of the technical solution provided by the present disclosure include at least:

[0051] It can be seen from the above technical solution that the embodiment of the present disclosure proposes a method for effectively verifying and subsequently processing the correctness of the target object detection program for the mining field. Specifically, by obtaining the perception data collected by the perception equipment of the unmanned vehicle, the perception data may include multiple frames of point cloud data corresponding to multiple different scenes, and then, for each frame of point cloud data corresponding to each scene, the target object true value calibration information of the frame point cloud data and the target object detection result of the frame point cloud data by the target object detection program are obtained, and the target object detection result of the frame point cloud data is matched with the target object true value calibration information of the frame point cloud data to obtain a matching processing result, and the target object missed detection rate, target object false detection rate and target detection error corresponding to the frame point cloud data are determined according to the matching processing result, and according to the target missed detection rate , the target object false detection rate and the target object detection error determine the test sub-result corresponding to the frame point cloud data, and the target test result corresponding to each scene is obtained according to the test sub-result of the multi-frame point cloud data corresponding to each scene, so that the processing strategy for the target object detection program can be determined according to the target test results corresponding to the multiple different scenes, and the detection program is processed according to the processing strategy. Since the target object detection program can be jointly tested based on point cloud data of multiple different scenes to obtain more comprehensive and effective test results, and the processing strategy for the target object detection program is determined based on the test results, the effectiveness of the correctness verification of the unmanned vehicle target object detection program is improved, and the reliability of the unmanned vehicle target object detection program can be further improved by processing the target object detection program according to the adapted processing strategy.

[0052] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0054] Figure 1 is a schematic diagram of a processing method for an object detection program of an unmanned vehicle provided by one embodiment of the present disclosure;

[0055] Figure 2 is a schematic diagram of a method for processing an object detection program for an unmanned vehicle provided by one embodiment of the present disclosure;

[0056] Figure 3This is a structural block diagram of a processing device for an object detection program for an unmanned vehicle provided by one embodiment of the present disclosure;

[0057] Figure 4 4 is a block diagram of an electronic device used to implement the processing method of the target object detection program of the unmanned vehicle of the embodiment of the present disclosure. DETAILED DESCRIPTION

[0058] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0059] Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present disclosure.

[0060] It should be noted that the terminal devices involved in the embodiments of the present disclosure may include but are not limited to mobile phones, personal digital assistants (PDAs), wireless handheld devices, tablet computers and other smart devices; display devices may include but are not limited to personal computers, televisions and other devices with display functions.

[0061] In this document, the term "and / or" simply describes a relationship between related objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " in this document generally indicates that the related objects are in an "or" relationship.

[0062] Please refer to Figure 1 , which shows a flow chart of a method for processing an object detection program for an unmanned vehicle provided by an embodiment of the present disclosure. The method for processing an object detection program for an unmanned vehicle may specifically include:

[0063] Step 101: Acquire perception data collected by the perception device of the unmanned vehicle, wherein the perception data includes multiple frames of point cloud data corresponding to multiple different scenes.

[0064] Step 102: For each frame of point cloud data corresponding to each scene, obtain the target object true value calibration information of the frame of point cloud data.

[0065] Step 103: Obtain the target object detection result of the target object detection program on the frame of point cloud data.

[0066] Step 104 : Match the target object detection result of the frame point cloud data with the target object true value calibration information of the frame point cloud data to obtain a matching result.

[0067] Step 105: Determine the target omission rate, target false detection rate, and target detection error corresponding to the frame point cloud data based on the matching processing result, and determine the test sub-result corresponding to the frame point cloud data based on the target omission rate, the target false detection rate, and the target detection error.

[0068] Step 106: Obtain a target test result corresponding to each scene based on the test sub-results of the multi-frame point cloud data corresponding to each scene.

[0069] Step 107: Determine a processing strategy for the target object detection program according to the target test results corresponding to the multiple different scenarios, and process the detection program according to the processing strategy.

[0070] It should be noted that the target missed detection rate can be used to represent the ratio of the target true value calibration information showing the existence of the target and the corresponding target detection result showing the absence of the corresponding target.

[0071] The target object false detection rate can be used to represent the ratio of the target object true value calibration information showing that the target object does not exist and the corresponding target object detection result showing that the corresponding target object exists.

[0072] The target detection error is used to characterize the distance error between two targets when the target true value calibration information shows that the target exists and the corresponding target detection result also shows that the corresponding target exists.

[0073] It should be noted that the target object true value calibration information can be obtained by pre-data calibration processing.

[0074] For example, a frame of point cloud data can be extracted from the collected perception data every 1 second, and each frame of point cloud data can be calibrated to obtain the target object's true value calibration information for each frame of point cloud data. Wherein, a frame of point cloud data can be a point cloud collected for a preset time duration (for example, 100 milliseconds).

[0075] It should be noted that, in step 107 , the target test results corresponding to the multiple different scenarios can be used to verify the correctness of the target object detection program, so as to determine the processing strategy for the target object detection program according to the verification results.

[0076] Among them, in an optional embodiment, the target object detection program can be processed in different ways according to the correctness verification results of the target test results corresponding to different scenarios. For example, if the target test result of the low-risk scenario has a high pass rate, while the target test result of the high-risk scenario has a low pass rate, the program will be updated / upgraded / corrected immediately; if the target test result of the low-risk scenario has a low pass rate, while the target test result of the high-risk scenario has a high pass rate, the program will be updated / upgraded / corrected and wait (pending further verification or execution of the next update node).

[0077] In another optional implementation, the target test results corresponding to different scenarios can be further processed to obtain a final target test result, and the program processing strategy can be determined based on the final target test result. For example, the test pass rates corresponding to the target test results corresponding to different scenarios can be averaged to obtain the final target test result. If the final target test result shows a high pass rate, the program is updated / upgraded / corrected and waits (pending further verification or execution at the next update node); otherwise, the program is updated / upgraded / corrected immediately.

[0078] In addition, based on the correctness verification results of the target test results corresponding to different scenarios, the target object detection program can also be updated / upgraded / corrected in different program modules. Furthermore, if the target object detection program is determined to require updating / upgrading / correction based on the target test results corresponding to multiple different scenarios, a prompt message can be output, and the program, test results, and test process can be sent to the update / upgrade / correction node. The update / upgrade / correction node then updates / upgrades / corrects the target object detection program based on the above information. It should be noted that the specific processing strategy can be set according to actual business needs.

[0079] It should be noted that part or all of the execution entities of steps 101 to 107 may be applications located in the local terminal, or may also be functional units such as plug-ins or software development kits (SDKs) set in the applications located in the local terminal, or may also be processing engines located in the network-side server, or may also be distributed systems located on the network side, for example, processing engines or distributed systems in the data testing platform on the network side, etc. This embodiment does not specifically limit this.

[0080] It is understandable that the application may be a native program (nativeApp) installed on the local terminal, or may be a webpage program (webApp) of a browser on the local terminal, which is not limited in this embodiment.

[0081] In this way, the target object detection program can be jointly tested based on point cloud data of multiple scenes to obtain more comprehensive and effective test results, and the processing strategy for the target object detection program can be determined based on the test results, thereby improving the effectiveness of the correctness verification of the unmanned vehicle target object detection program. By processing the target object detection program according to the adapted processing strategy, the reliability of the unmanned vehicle target object detection program can be further improved.

[0082] Optionally, in a possible implementation of this embodiment, in step 105, specifically when the target object omission detection rate is less than a preset first threshold, the target object false detection rate is less than a preset second threshold, and the target object detection error is less than a preset third threshold, the test sub-result can be determined as the target object detection program's perception ability for the frame point cloud data has passed the test; and when the target object omission detection rate is greater than or equal to the preset first threshold, the target object false detection rate is greater than or equal to the preset second threshold, and / or the target object detection error is greater than or equal to the preset third threshold, the test sub-result can be determined as the target object detection program's perception ability for the frame point cloud data has failed the test.

[0083] In this implementation, the preset first threshold may be a target missed detection rate threshold determined based on actual business scenarios. The preset second threshold may be a target false detection rate threshold determined based on actual business scenarios. The preset third threshold may be a target detection error threshold determined based on actual business scenarios, or a target detection error threshold determined based on the average error of target detection in each frame.

[0084] Preferably, for a single frame of point cloud data, the preset first threshold may be 20%; the preset second threshold may be 20%; and the preset third threshold may be the average error of target object detection in the frame.

[0085] In a specific implementation process of this implementation method, for each frame of point cloud data corresponding to each scene, after obtaining the test sub-result of each frame of point cloud data, the test pass rate of the point cloud data corresponding to each scene can be calculated.

[0086] In this way, by comparing the target object missed detection rate, target object false detection rate, and target object detection error with the corresponding preset thresholds, more accurate and effective test sub-results for each frame of point cloud data corresponding to each scene can be obtained, so that more accurate and effective target test results for each scene can be obtained later, thereby further improving the effectiveness of the correctness verification of the test results and further improving the reliability of the target object detection program of the unmanned vehicle.

[0087] Optionally, in a possible implementation of this embodiment, in step 106, the test pass rate of the point cloud data corresponding to the scene can be obtained based on the test sub-results as the target test result corresponding to the scene.

[0088] In a specific implementation process of this implementation method, after obtaining the target test results corresponding to each scene, in step 107, the correction coefficients corresponding to the multiple different scenes can be obtained, and then the final test pass rate can be calculated based on the test pass rates corresponding to the multiple different scenes and the correction coefficients corresponding to the multiple different scenes, so that the processing strategy for the target object detection program can be determined based on the final test pass rate.

[0089] In one case of the specific implementation process, a weighted average calculation process can be performed based on the test pass rates corresponding to multiple different scenarios and the correction coefficients corresponding to multiple different scenarios to obtain the final test pass rate of the target object detection program.

[0090] Specifically, the test pass rates corresponding to different scenarios can be multiplied by their corresponding correction coefficients, and then the sum of each product can be calculated and divided by the number of scenarios to obtain the final test pass rate of the target object detection program.

[0091] In this implementation, the magnitude of the correction coefficient is related to the danger level of the scene. Specifically, the higher the danger level of the scene, the smaller the correction coefficient. The correction coefficient is less than or equal to 1.

[0092] It is understandable that for less dangerous scenarios, such as mining roads with low traffic density, the test pass rate requirement can be more relaxed, that is, a relatively large correction factor can be pre-configured for this scenario. For more dangerous scenarios, such as loading areas with more intensive work activities, the test pass rate requirement is more stringent, that is, a smaller correction factor can be pre-configured for this scenario.

[0093] In this way, the final test pass rate of the target object detection program can be determined by combining the correction coefficient that represents the degree of danger, which can more effectively verify the target object detection program, thereby further improving the effectiveness and reliability of the correctness verification of the unmanned vehicle target object detection program.

[0094] The above-mentioned different scenarios can be different types of working areas, different types of obstacles, different driving parameters, etc. For example, in the case of different driving parameters, the driving parameter is the driving speed, including 5km / h, 10km / h, and 15km / h. The test vehicle drives towards the vehicle in front at three driving speeds respectively. The collection ends when the tails of the two vehicles are 10m apart. The pass rates of the three cases are A, B, and C respectively. According to the above embodiment, the final test pass rate = (1*A+0.9*B+0.8*C) / 3, where the correction coefficients corresponding to the three cases are 1, 0.9, and 0.8 respectively. It should be noted that the correction coefficient can be specified according to the actual business needs, and 1, 0.9, and 0.8 are only examples.

[0095] It is understandable that other existing methods can also be used to calculate the final test pass rate, for example, directly calculating the average of the pass rates of various scenarios, which will not be described in detail here.

[0096] In another specific implementation of this method, for any scene's point cloud data, if the object detection program's test pass rate exceeds a preset pass rate threshold, the object detection program for that scene can be determined to have passed the test. If the number of scenes that passed the test does not meet the preset threshold, the object detection program can be adjusted and retested. If the object detection program passes the test for each scene, the object detection program can be used for actual prediction.

[0097] It should be noted that the specific implementation process provided in this implementation can be combined with the various specific implementation processes provided in the aforementioned implementations to implement the processing method of the target object detection program for the unmanned vehicle of this embodiment. A detailed description can be found in the relevant content of the aforementioned implementations and will not be repeated here.

[0098] Optionally, in a possible implementation of this embodiment, in step 104, the intersection-and-union ratio of the target object detection result of the frame point cloud data and the target object true value calibration information of the frame point cloud data can be specifically determined as the matching processing result.

[0099] In this implementation, the target object detection result of the frame point cloud data may include the size and center position information of the target detection frame, and the target object true value calibration information of the frame point cloud data may include the size and center position of the target object true value calibration detection frame.

[0100] Specifically, the size and center position of the target detection frame can be calibrated based on the target's true value, and the intersection-over-union (IoU) of the point cloud data for that frame can be calculated. The IoU is equal to the intersection of the true value and the detection result divided by the union of the true value and the detection result.

[0101] In this implementation, after obtaining the matching processing result, in step 105, the target object detection error, target object missed detection rate and target object false detection rate corresponding to the frame point cloud data can be obtained respectively through the following specific implementation process.

[0102] In a specific implementation of this method, specifically, first, the true value calibration information of the target object that is successfully matched in the frame of point cloud data can be determined based on the intersection-over-union ratio of the frame of point cloud data and a preset fourth threshold. Second, the distance difference between the true value calibration information of the target object that is successfully matched in the frame of point cloud data and the corresponding target object detection result is calculated to obtain the target object detection error corresponding to the frame of point cloud data.

[0103] Specifically, the preset fourth threshold may be a preset intersection-over-union ratio threshold. Preferably, the preset fourth threshold may be 0.6.

[0104] In this implementation, for a target object true value calibration information, if the intersection-and-union ratio of the target object true value calibration information and a target test result is greater than the preset fourth threshold, the target object true value calibration information can be matched successfully; if the intersection-and-union ratio of the target object true value calibration information and a target test result is less than or equal to the preset fourth threshold, the target object true value calibration information cannot be matched successfully.

[0105] Optionally, corresponding to any successfully matched target object true value calibration information, the center position of the successfully matched target object true value calibration information in the frame point cloud data and the center position of the corresponding target object detection result can be obtained, and the Euclidean distance between the center position of the successfully matched target object true value calibration information in the frame point cloud data and the center position of the corresponding target object detection result can be calculated to obtain the target object detection error.

[0106] Furthermore, the average error of the frame of point cloud data may be calculated based on all successfully matched target object detection errors and the number of successfully matched target objects in the frame of point cloud data.

[0107] In another specific implementation of this method, specifically, first, the unmatched target object true value calibration information in the frame of point cloud data can be determined based on the intersection-over-union ratio of the frame of point cloud data and a preset fourth threshold. Second, the number of unmatched target object true value calibration information in the frame of point cloud data and the total number of target object true value calibration information in the frame of point cloud data are determined. Third, the ratio between the number of unmatched target object true value calibration information and the total number of target object true value calibration information is used as the target object missed detection rate.

[0108] In another specific implementation of this method, specifically, first, unmatched target object detection results in the frame of point cloud data can be determined based on the intersection-over-union ratio of the frame of point cloud data and a preset fourth threshold. Second, the number of unmatched target object detection results in the frame of point cloud data and the total number of target object detection results in the frame of point cloud data are determined. Third, the ratio between the number of unmatched target object detection results and the total number of target object detection results is used as the target object false detection rate.

[0109] It can be understood that here, if the intersection-and-union ratio of a target object's true value calibration information and the target detection result does not meet the preset fourth threshold, that is, it fails to match the target detection result, then the target object is judged to be missed; if a target object's true value calibration information only successfully matches one target detection result, the missed detection rate can continue to be calculated; if a target object's true value calibration information successfully matches two or more target detection results, the target detection result with a larger intersection-and-union ratio can be selected; after traversing all target object true value calibration information, the target detection result that has not been successfully matched is a false detection.

[0110] In this way, the accuracy and effectiveness of the evaluation of the test program can be improved by determining the intersection-over-union ratio of the target object detection result of the frame point cloud data and the target object true value calibration information of the frame point cloud data, and using the intersection-over-union ratio as the comparison standard for evaluating the target detection result of the test program.

[0111] It should be noted that the specific implementation process provided in this implementation can be combined with the various specific implementation processes provided in the aforementioned implementations to implement the processing method of the target object detection program for the unmanned vehicle of this embodiment. A detailed description can be found in the relevant content of the aforementioned implementations and will not be repeated here.

[0112] Optionally, in a possible implementation of this embodiment, the multiple different scenarios include test scenarios corresponding to at least two different types of working areas in a mining area, or the multiple different scenarios include test scenarios corresponding to at least two types of obstacles.

[0113] Optionally, the multiple different scenarios may include test scenarios corresponding to at least two different types of working areas in a mining area, and the multiple different scenarios may include test scenarios corresponding to at least two types of obstacles.

[0114] In this implementation, the working area of ​​the mining area may include but is not limited to a loading area, a transport road, a spoil dump, a parking lot, etc. The transport road may include an uphill road and a downhill road.

[0115] In this implementation, obstacle types include, but are not limited to, static and dynamic obstacles. Static obstacles include, but are not limited to, rocks, ruts, retaining walls, parked vehicles, and other static structures in the mining area. Dynamic obstacles include, but are not limited to, vehicles / individuals traveling in the same direction and vehicles / individuals traveling in opposite directions. Vehicles include mining trucks, command vehicles, and forklifts.

[0116] In a specific implementation process of this method, multiple stones with a certain horizontal distance between them can be placed at a preset first distance in front of the unmanned vehicle that collects perception data in loading areas, uphill roads, downhill roads, spoil dumps, and parking lots. The unmanned vehicle can accelerate from a standstill to multiple different speeds and then travel at a constant speed until it stops at a preset second distance from the stones, completing the perception data collection.

[0117] For example, in loading areas, uphill roads, downhill roads, spoil dumps, and parking lots, two stones 3 meters apart can be placed 100 meters in front of the unmanned vehicle that collects perception data. The vehicle accelerates from a standstill to 5 km / h, 10 km / h, 20 km / h, 25 km / h, and 30 km / h, respectively, and then drives at a constant speed until it stops 3 meters away from the stones to complete data collection.

[0118] In another specific implementation process of this implementation method, the unmanned vehicle that collects perception data can maintain multiple different speeds for loading areas, roads, spoil dumps, and parking lots, and oncoming vehicles can maintain multiple different speeds until the two vehicles are at a preset third distance apart at the rear, completing the perception data collection.

[0119] For example, for the loading area, road, spoil dump, and parking lot scenes, the unmanned vehicle collecting perception data remains stationary at 0 km / h, 5 km / h, 10 km / h, 20 km / h, 25 km / h, and 30 km / h, and the oncoming vehicles maintain a speed of 5 km / h, 10 km / h, 20 km / h, 25 km / h, and 30 km / h respectively, until the two vehicles are 10 meters apart at the rear, completing the perception data collection.

[0120] In another specific implementation process of this implementation method, the collected unmanned vehicle can start from a standstill for scenes such as loading areas, roads, spoil dumps, and parking lots. At a preset fourth distance in front of the collected unmanned vehicle, there is a vehicle that maintains multiple speeds and travels in the same direction. The collected unmanned vehicle accelerates to multiple different speeds and then travels at a constant speed, collecting perception data according to the preset time length.

[0121] For example, for scenes such as loading areas, roads, spoil dumps, and parking lots, the unmanned vehicle collecting data starts from a standstill. 100 meters in front of the unmanned vehicle collecting perception data, there is a car driving in the same direction at speeds of 5 km / h, 10 km / h, 20 km / h, 25 km / h, and 30 km / h. The unmanned vehicle collecting data accelerates to 5 km / h, 10 km / h, 20 km / h, 25 km / h, and 30 km / h respectively, and then drives at a constant speed to collect perception data for one minute.

[0122] In another specific implementation process of this implementation method, cones can be placed at a preset fifth distance in front of the unmanned vehicle that collects perception data to simulate pedestrians in loading areas, roads, spoil dumps, and parking lots. The collecting unmanned vehicle starts from a standstill, accelerates to multiple different speeds, and then drives at a constant speed until it stops at a preset sixth distance from the cones, completing the perception data collection.

[0123] For example, in loading areas, roads, spoil dumps, and parking lots, cones can be placed 100 meters in front of the unmanned vehicle to simulate pedestrians. The unmanned vehicle collecting perception data starts from a standstill and accelerates to 5 km / h, 10 km / h, 20 km / h, 25 km / h, and 30 km / h, respectively, and then drives at a constant speed until it stops 3 meters away from the cones, completing the perception data collection.

[0124] It is understood that the oncoming and same-direction vehicles can be replaced with mining trucks, command vehicles, and shovels for data collection, respectively. These vehicles can be manned or unmanned. The vehicles used for data collection can also be unmanned mining trucks, unmanned command vehicles, and unmanned shovels.

[0125] In this way, by collecting perception data from multiple different scenarios for program testing, the comprehensiveness and richness of the test scenarios can be improved, while also enhancing the targetedness of the test data for each scenario. This allows for joint testing based on the perception data of these scenarios, further improving the effectiveness and reliability of the correctness verification of the target object detection program.

[0126] It should be noted that the specific implementation process provided in this implementation can be combined with the various specific implementation processes provided in the aforementioned implementations to implement the processing method of the target object detection program for the unmanned vehicle of this embodiment. A detailed description can be found in the relevant content of the aforementioned implementations and will not be repeated here.

[0127] In order to better understand the method of the embodiment of the present application, the method of the embodiment of the present application is described below with reference to the accompanying drawings and specific mining application scenarios.

[0128] Figure 2FIG. 1 is a flow chart of a method for processing an object detection program for an unmanned vehicle provided by another embodiment of the present application. Figure 2 shown.

[0129] Step 201: Obtain multi-frame point cloud data corresponding to multiple different scenes in the mining area collected by the perception equipment of the unmanned vehicle.

[0130] In this embodiment, the multiple different scenarios may include test scenarios corresponding to multiple different types of operating areas in a mining area, or the multiple different scenarios may include test scenarios corresponding to at least two types of obstacles.

[0131] Optionally, the multiple different scenarios may be test scenarios corresponding to multiple different types of operating areas, or may be test scenarios corresponding to multiple types of obstacles.

[0132] For example, the three different scenarios may be a scenario where multiple vehicles are working in a staggered manner in a loading area, a scenario where a transport road has stone roadblocks and pedestrians, a scenario where multiple vehicles are working in a staggered manner in a spoil dump, etc.

[0133] It is understandable that the specific method of collecting data for a scene can be found in the relevant content of the aforementioned embodiment and will not be described in detail here.

[0134] Step 202: For each frame of point cloud data corresponding to each scene, obtain the target object true value calibration information of the frame of point cloud data and the target object detection result of the target object detection program for the frame of point cloud data.

[0135] Step 203: Determine an intersection-over-union ratio between the target object detection result of the frame point cloud data and the target object true value calibration information of the frame point cloud data as a matching processing result.

[0136] Step 204 : Determine the true value calibration information of the target object that has been successfully matched in the frame of point cloud data based on the intersection-over-union ratio of the frame of point cloud data and a preset intersection-over-union ratio threshold.

[0137] Step 205: Calculate the distance difference between the true value calibration information of the successfully matched target object in the frame point cloud data and the corresponding target object detection result to obtain the target object detection error corresponding to the frame point cloud data.

[0138] Step 206 : Determine the true value calibration information of the target object that is not successfully matched in the frame of point cloud data according to the intersection-over-union ratio of the frame of point cloud data and a preset intersection-over-union ratio threshold.

[0139] Step 207: Determine the number of target object true value calibration information that has not been successfully matched in the frame of point cloud data and the total number of target object true value calibration information in the frame of point cloud data.

[0140] Step 208: The ratio between the number of target object true value calibration information that was not successfully matched and the total number of target object true value calibration information is used as the target object missed detection rate.

[0141] Step 209 : Determine the detection result of the target object that is not successfully matched in the frame of point cloud data according to the IoU ratio of the frame of point cloud data and a preset IoU ratio threshold.

[0142] Step 210: Determine the number of unmatched target object detection results in the frame of point cloud data and the total number of target object detection results in the frame of point cloud data.

[0143] Step 211: The ratio between the number of unmatched target object detection results and the total number of target object detection results is used as the target object false detection rate.

[0144] In this embodiment, for a target object true value calibration information, if the intersection-and-union ratio of the target object true value calibration information and a target test result is greater than the preset intersection-and-union ratio threshold, the target object true value calibration information can be matched successfully; if the intersection-and-union ratio of the target object true value calibration information and a target test result is less than or equal to the preset intersection-and-union ratio threshold, the target object true value calibration information can be matched unsuccessfully.

[0145] Step 212: When the target object missed detection rate is less than the preset first threshold, the target object false detection rate is less than the preset second threshold, and the target object detection error is less than the preset third threshold, determine that the test sub-result is that the target object detection program's perception ability of the frame point cloud data has passed the test.

[0146] Step 213: When the target object missed detection rate is greater than or equal to the preset first threshold, the target object false detection rate is greater than or equal to the preset second threshold, and / or the target object detection error is greater than or equal to the preset third threshold, determine that the test sub-result is that the target object detection program's perception ability of the frame point cloud data has failed the test.

[0147] Step 214: According to the test sub-results, obtain the test pass rate of the point cloud data corresponding to the scene as the target test result corresponding to the scene.

[0148] In this embodiment, the test sub-results of a single frame of point cloud data may include that the object detection program's perception capability of the frame of point cloud data failed the test, and that the object detection program's perception capability of the frame of point cloud data passed the test.

[0149] Specifically, each scene may include multiple frames of point cloud data. By processing each frame of point cloud data from steps 202 to 213, a test sub-result of each frame of point cloud data can be obtained. Then, based on the test sub-result of each frame of point cloud data, that is, whether the test was passed or failed, the overall test pass rate of the scene can be calculated, and the pass rate can be used as the target test result corresponding to the scene.

[0150] Step 215: Obtain correction coefficients corresponding to multiple different scenes.

[0151] Step 216: Calculate a final test pass rate based on the test pass rates corresponding to the multiple different scenarios and the correction coefficients corresponding to the multiple different scenarios.

[0152] In this embodiment, the magnitude of the correction coefficient is related to the degree of danger corresponding to the scene.

[0153] Optionally, a weighted average calculation may be performed on the test pass rates corresponding to a plurality of different scenarios and the correction coefficients corresponding to the plurality of different scenarios to obtain a final test pass rate.

[0154] For example, if the different scenarios include scenario A, scenario B, and scenario C, the pass rate of scenario A is x, corresponding to the correction coefficient p, the pass rate of scenario B is y, corresponding to the correction coefficient m, and the pass rate of scenario C is z, corresponding to the correction coefficient n, the final test pass rate Q can be calculated by formula (1):

[0155] Q=(x*p+y*m+z*n) / 3 (1)

[0156] Step 217: Determine a processing strategy for the target object detection program based on the final test pass rate.

[0157] At this point, after the final test pass rate of the object detection program is determined, it can be determined based on the final test pass rate whether to adjust the object detection program or whether to apply the object detection program to an actual object detection task.

[0158] In this embodiment, the target object detection program can be jointly tested based on point cloud data of multiple scenes to obtain more comprehensive and effective test results, and the processing strategy for the target object detection program can be determined based on the test results, thereby improving the effectiveness of the correctness verification of the unmanned vehicle target object detection program. By processing the target object detection program according to the adapted processing strategy, the reliability of the unmanned vehicle target object detection program can be further improved.

[0159] Moreover, by determining the intersection-over-union ratio of the target object detection results of each frame of point cloud data in each scene and the target object true value calibration information of the frame of point cloud data, the intersection-over-union ratio can be used as a comparison standard for evaluating the target detection results of the test program, which can improve the accuracy and effectiveness of the evaluation of the test program.

[0160] It should be noted that for the aforementioned method embodiments, for simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present disclosure is not limited by the order of the actions described, because according to the present disclosure, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present disclosure.

[0161] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0162] Figure 3 FIG. 1 shows a structural block diagram of a processing device for an object detection program for an unmanned vehicle provided by an embodiment of the present disclosure. Figure 3 As shown. The processing 300 of the target object detection program of the unmanned vehicle of this embodiment may include a first acquisition unit 301, a second acquisition unit 302, a third acquisition unit 303, a first matching unit 304, a first determination unit 305, a first acquisition unit 306 and a second determination unit 307. Among them, the first acquisition unit 301 is used to obtain the perception data collected by the perception device of the unmanned vehicle, and the perception data includes multiple frames of point cloud data corresponding to multiple different scenes; for each frame of point cloud data corresponding to each scene, the second acquisition unit 302 is used to obtain the target object true value calibration information of the frame point cloud data; the third acquisition unit 303 is used to obtain the target object detection result of the frame point cloud data by the target object detection program; the first matching unit 304 is used to match the target object detection result of the frame point cloud data with the target object true value calibration information of the frame point cloud data to obtain the matching processing result; the first determination unit 305 is used to The target omission rate, target false detection rate and target detection error corresponding to the frame point cloud data are determined according to the matching processing results, and the test sub-result corresponding to the frame point cloud data is determined according to the target omission rate, the target false detection rate and the target detection error; a first obtaining unit 306 is used to obtain the target test result corresponding to each scene according to the test sub-results of the multiple frames of point cloud data corresponding to each scene; a second determining unit 307 is used to determine the processing strategy for the target detection program according to the target test results corresponding to the multiple different scenes, and process the detection program according to the processing strategy.

[0163] It should be noted that part or all of the heading angle determination device of the laser radar of this embodiment can be an application located in the local terminal, or it can also be a functional unit such as a plug-in or software development kit (SDK) set in the application located in the local terminal, or it can also be a processing engine located in the network side server, or it can also be a distributed system located on the network side, for example, a processing engine or distributed system in the data testing platform on the network side, etc. This embodiment does not specifically limit this.

[0164] It is understandable that the application may be a native program (nativeApp) installed on the local terminal, or may be a webpage program (webApp) of a browser on the local terminal, which is not limited in this embodiment.

[0165] Optionally, in a possible implementation of this embodiment, the first determination unit 305 can be specifically used to determine that the test sub-result is that the target object detection program's perception ability for the frame point cloud data has passed the test when the target object omission detection rate is less than a preset first threshold, the target object false detection rate is less than a preset second threshold, and the target object detection error is less than a preset third threshold; and to determine that the test sub-result is that the target object detection program's perception ability for the frame point cloud data has failed the test when the target object omission detection rate is greater than or equal to the preset first threshold, the target object false detection rate is greater than or equal to the preset second threshold, and / or the target object detection error is greater than or equal to the preset third threshold.

[0166] Optionally, in a possible implementation of this embodiment, the first obtaining unit 306 may be specifically configured to obtain a test pass rate of the point cloud data corresponding to the scene according to the test sub-results, as a target test result corresponding to the scene.

[0167] Optionally, in a possible implementation of this embodiment, the second determination unit 307 can be specifically used to obtain correction coefficients corresponding to the multiple different scenarios, where the size of the correction coefficient is related to the degree of danger corresponding to the scenario. The final test pass rate is calculated based on the test pass rates corresponding to the multiple different scenarios and the correction coefficients corresponding to the multiple different scenarios. The processing strategy for the target object detection program is determined based on the final test pass rate.

[0168] Optionally, in a possible implementation of this embodiment, the first matching unit 304 can be specifically used to determine the intersection ratio of the target detection result of the frame point cloud data and the target true value calibration information of the frame point cloud data as the matching processing result.

[0169] Optionally, in a possible implementation of this embodiment, the first determination unit 305 can be specifically used to determine the target object detection error corresponding to the frame point cloud data according to the matching processing result in the following manner: according to the intersection-over-union ratio of the frame point cloud data and a preset fourth threshold, determine the true value calibration information of the target object that is successfully matched in the frame point cloud data, calculate the distance difference between the true value calibration information of the target object that is successfully matched in the frame point cloud data and the corresponding target object detection result, and obtain the target object detection error corresponding to the frame point cloud data.

[0170] Optionally, in a possible implementation of this embodiment, the first determination unit 305 can be specifically used to determine the target object omission rate corresponding to the frame point cloud data according to the matching processing result in the following manner: determine the target object true value calibration information that has not been successfully matched in the frame point cloud data according to the intersection-union ratio of the frame point cloud data and a preset fourth threshold, determine the number of target object true value calibration information that has not been successfully matched in the frame point cloud data and the total number of target object true value calibration information in the frame point cloud data, and use the ratio between the number of target object true value calibration information that has not been successfully matched and the total number of target object true value calibration information as the target object omission rate.

[0171] Optionally, in a possible implementation of this embodiment, the first determination unit 305 can be specifically used to determine the target object false detection rate corresponding to the frame point cloud data according to the matching processing result in the following manner: determine the target object detection results that were not successfully matched in the frame point cloud data according to the intersection-union ratio of the frame point cloud data and a preset fourth threshold, determine the number of target object detection results that were not successfully matched in the frame point cloud data and the total number of target object detection results in the frame point cloud data, and use the ratio between the number of target object detection results that were not successfully matched and the total number of target object detection results as the target object false detection rate.

[0172] Optionally, in a possible implementation of this embodiment, the multiple different scenarios include test scenarios corresponding to at least two different types of working areas in a mining area; and / or, the multiple different scenarios include test scenarios corresponding to at least two types of obstacles.

[0173] In this embodiment, the perception data collected by the perception device of the unmanned vehicle is obtained by the first acquisition unit, and the perception data includes multiple frames of point cloud data corresponding to multiple different scenes. Then, for each frame of point cloud data corresponding to each scene, the second acquisition unit can obtain the target object true value calibration information of the frame point cloud data, and the third acquisition unit can obtain the target object detection result of the frame point cloud data by the target object detection program. The first matching unit matches the target object detection result of the frame point cloud data with the target object true value calibration information of the frame point cloud data to obtain a matching processing result. The first determination unit determines the target object missed detection rate, target object false detection rate and target object detection error corresponding to the frame point cloud data according to the matching processing result, and determines the target object missed detection rate, target object false detection rate and target object detection error according to the target missed detection rate, the target object false detection rate and the target object detection error. The test sub-result corresponding to the frame point cloud data is determined by the difference, and the first obtaining unit obtains the target test result corresponding to each scene according to the test sub-result of the multi-frame point cloud data corresponding to each scene, so that the second determining unit can determine the processing strategy for the target object detection program according to the target test results corresponding to the multiple different scenes, and process the detection program according to the processing strategy. Since the target object detection program can be jointly tested based on point cloud data of multiple scenes to obtain more comprehensive and effective test results, and the processing strategy for the target object detection program is determined based on the test results, the effectiveness of the correctness verification of the unmanned vehicle target object detection program is improved, and the reliability of the unmanned vehicle target object detection program can be further improved by processing the target object detection program according to the adapted processing strategy.

[0174] In the technical solutions disclosed herein, the collection, storage, use, processing, transmission, provision and disclosure of user personal information, such as user images and attribute data, comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0175] According to an embodiment of the present application, the present application also provides an electronic device, a readable storage medium and a computer program product.

[0176] According to an embodiment of the present application, further, an unmanned vehicle including the provided electronic device is provided, and the autonomous driving vehicle may include an L2 and above level unmanned driving vehicle.

[0177] Figure 4A schematic block diagram of an example electronic device 400 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are provided as examples only and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0178] like Figure 4 As shown, electronic device 400 includes a computing unit 401, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 402 or a computer program loaded from a storage unit 408 into a random access memory (RAM) 403. Various programs and data required for the operation of electronic device 400 can also be stored in RAM 403. Computing unit 401, ROM 402, and RAM 403 are connected to each other via a bus 404. An input / output (I / O) interface 405 is also connected to bus 404.

[0179] Multiple components in the electronic device 400 are connected to the I / O interface 405, including an input unit 406, such as a keyboard, a mouse, etc.; an output unit 407, such as various types of displays, speakers, etc.; a storage unit 408, such as a magnetic disk, an optical disk, etc.; and a communication unit 409, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 409 allows the electronic device 400 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0180] The computing unit 401 can be a variety of general-purpose and / or specialized processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 401 performs the various methods and processes described above, such as the processing method of the object detection program of the unmanned vehicle. For example, in some embodiments, the processing method of the object detection program of the unmanned vehicle can be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as the storage unit 408. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 400 via the ROM 402 and / or the communication unit 409. When the computer program is loaded into the RAM 403 and executed by the computing unit 401, one or more steps of the processing method of the object detection program of the unmanned vehicle described above can be performed. Alternatively, in other embodiments, the computing unit 401 may be configured to execute the processing method of the target object detection program of the unmanned vehicle in any other appropriate manner (e.g., by means of firmware).

[0181] Various embodiments of the systems and techniques described above can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0182] The program code for implementing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0183] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0184] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0185] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0186] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact through a communication network. The client-server relationship arises through computer programs running on the respective computers and having a client-server relationship with each other. The server may be a cloud server, a server in a distributed system, or a server integrated with a blockchain.

[0187] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not a limitation herein.

[0188] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A method for processing an object detection program of an unmanned vehicle, characterized in that: The method comprises: Acquire perception data collected by the perception device of the unmanned vehicle, wherein the perception data includes multiple frames of point cloud data corresponding to multiple different scenes; For each frame of point cloud data corresponding to each scene, perform the following operations: Obtain the target object true value calibration information of the frame point cloud data; Obtain the target object detection result of the target object detection program for the frame point cloud data; Matching the target object detection result of the frame point cloud data with the target object true value calibration information of the frame point cloud data to obtain a matching processing result; Determine, based on the matching processing result, a target object missed detection rate, a target object false detection rate, and a target object detection error corresponding to the frame point cloud data, and determine a test sub-result corresponding to the frame point cloud data based on the target object missed detection rate, the target object false detection rate, and the target object detection error; Obtain the target test result corresponding to each scene based on the test sub-results of the multi-frame point cloud data corresponding to each scene; Based on the target test results corresponding to the multiple different scenarios, a processing strategy for the target object detection program is determined, and the detection program is processed according to the processing strategy, including: obtaining correction coefficients corresponding to the multiple different scenarios, the size of the correction coefficient is related to the degree of danger corresponding to the scenario; based on the test pass rates corresponding to the multiple different scenarios and the correction coefficients corresponding to the multiple different scenarios, a final test pass rate is calculated; based on the final test pass rate, a processing strategy for the target object detection program is determined.

2. The method according to claim 1, characterized in that The determining of a test sub-result corresponding to the frame of point cloud data according to the target object missed detection rate, the target object false detection rate, and the target object detection error includes: If the target object missed detection rate is less than a preset first threshold, the target object false detection rate is less than a preset second threshold, and the target object detection error is less than a preset third threshold, determining the test sub-result as passing the test on the perception capability of the target object detection program for the frame of point cloud data; When the target object missed detection rate is greater than or equal to the preset first threshold, the target object false detection rate is greater than or equal to the preset second threshold, and / or the target object detection error is greater than or equal to the preset third threshold, the test sub-result is determined to be that the target object detection program's perception ability of the frame point cloud data has failed the test.

3. The method according to claim 1 or 2, characterized in that Obtaining the target test result corresponding to each scene based on the test sub-results of the multi-frame point cloud data corresponding to each scene includes: According to the test sub-results, the test pass rate of the point cloud data corresponding to the scene is obtained as the target test result corresponding to the scene.

4. The method according to claim 1, wherein The matching process of the target object detection result of the frame point cloud data with the target object true value calibration information of the frame point cloud data to obtain a matching process result includes: An intersection-over-union ratio of the target object detection result of the frame point cloud data and the target object true value calibration information of the frame point cloud data is determined as the matching processing result.

5. The method according to claim 4, characterized in that The target object detection error corresponding to the frame point cloud data is determined according to the matching processing result in the following manner: Determining true value calibration information of the target object that successfully matched in the frame of point cloud data based on the intersection-over-union ratio of the frame of point cloud data and a preset fourth threshold; The distance difference between the true value calibration information of the target object that is successfully matched in the frame point cloud data and the corresponding target object detection result is calculated to obtain the target object detection error corresponding to the frame point cloud data.

6. The method according to claim 4, characterized in that The target object missed detection rate corresponding to the frame point cloud data is determined according to the matching processing result in the following manner: Determining true value calibration information of the target object that was not successfully matched in the frame of point cloud data based on the intersection-over-union ratio of the frame of point cloud data and a preset fourth threshold; Determine the number of target object true value calibration information that has not been successfully matched in the frame of point cloud data and the total number of target object true value calibration information in the frame of point cloud data; The ratio between the number of the target object true value calibration information that was not successfully matched and the total number of the target object true value calibration information is used as the target object missed detection rate.

7. The method according to claim 4, characterized in that The target object false detection rate corresponding to the frame point cloud data is determined according to the matching processing result in the following manner: Determining a detection result of an unsuccessful target object in the frame of point cloud data based on an intersection-over-union ratio of the frame of point cloud data and a preset fourth threshold; Determine the number of unmatched target object detection results in the frame of point cloud data and the total number of target object detection results in the frame of point cloud data; The ratio between the number of the unmatched target object detection results and the total number of the target object detection results is used as the target object false detection rate.

8. The method according to any one of claims 1, 2, 4 to 7, characterized in that The multiple different scenarios include test scenarios corresponding to at least two different types of operating areas in the mining area; and / or, The multiple different scenarios include test scenarios corresponding to at least two obstacle types.

9. A processing device for an object detection program of an unmanned vehicle, characterized in that: The device comprises: The first acquisition unit is used to acquire the perception data collected by the perception device of the unmanned vehicle, wherein the perception data includes multiple frames of point cloud data corresponding to multiple different scenes; for each frame of point cloud data corresponding to each scene, A second acquisition unit is used to obtain the target object true value calibration information of the frame point cloud data; A third acquisition unit is used to obtain the target object detection result of the target object detection program on the frame point cloud data; a first matching unit, configured to match the target object detection result of the frame point cloud data with the target object true value calibration information of the frame point cloud data to obtain a matching result; a first determining unit, configured to determine a target object missed detection rate, a target object false detection rate, and a target object detection error corresponding to the frame point cloud data based on the matching processing result, and determine a test sub-result corresponding to the frame point cloud data based on the target object missed detection rate, the target object false detection rate, and the target object detection error; A first obtaining unit is configured to obtain a target test result corresponding to each scene according to the test sub-results of the multi-frame point cloud data corresponding to each scene; The second determining unit is configured to determine a processing strategy for the target object detection program according to the target test results corresponding to the multiple different scenarios, and process the detection program according to the processing strategy, wherein: The second determination unit is also used to obtain the correction coefficients corresponding to the multiple different scenarios, the size of the correction coefficients is related to the degree of danger corresponding to the scenario; based on the test pass rates corresponding to the multiple different scenarios and the correction coefficients corresponding to the multiple different scenarios, a final test pass rate is calculated; based on the final test pass rate, a processing strategy for the target object detection program is determined.

Images