Perception evaluation method, device, and electronic device

Abstract

The present disclosure provides a perception assessment method, apparatus, electronic device, and storage medium, relating to the field of artificial intelligence, particularly autonomous driving, intelligent transportation, and other fields. A specific implementation scheme comprises: obtaining a vehicle's detection results of a target obstacle in multiple dimensions; wherein the detection results of the multiple dimensions are obtained by the vehicle during driving; and based on the detection results of the multiple dimensions and preset information about the target obstacle, obtaining a perception assessment result of the vehicle in each of the multiple dimensions. Using this scheme, an accurate and comprehensive assessment of the vehicle's obstacle detection results can be performed.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence, and in particular to the fields of autonomous driving, intelligent transportation, etc. Background Art

[0002] With the advancement of computer technology, artificial intelligence (AI) fields such as intelligent transportation have also experienced rapid growth. In particular, technologies such as autonomous vehicles are gaining widespread application. Autonomous driving requires the ability to safely and reliably detect obstacles to support safe operation. However, accurately and comprehensively evaluating the vehicle's obstacle detection results has become a challenge. Summary of the Invention

[0003] The present disclosure provides a perception evaluation method, device, electronic device, and storage medium.

[0004] According to a first aspect of the present disclosure, there is provided a perception assessment method, comprising:

[0005] Obtaining detection results of the vehicle on a target obstacle in multiple dimensions; wherein the detection results in multiple dimensions are obtained by the vehicle during driving;

[0006] Based on the detection results of the multiple dimensions and the preset information of the target obstacle, a perception evaluation result of the vehicle in each of the multiple dimensions is obtained.

[0007] According to a second aspect of the present disclosure, there is provided a perception assessment device, comprising:

[0008] An acquisition module is used to obtain the vehicle's detection results of multiple dimensions of the target obstacle; wherein the detection results of the multiple dimensions are obtained by the vehicle during driving;

[0009] An evaluation module is used to obtain a perception evaluation result of the vehicle in each of the multiple dimensions based on the detection results of the multiple dimensions and preset information of the target obstacle.

[0010] According to a third aspect of the present disclosure, there is provided an electronic device, including:

[0011] at least one processor; and

[0012] a memory communicatively connected to the at least one processor; wherein,

[0013] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the ground element information acquisition method of the first aspect mentioned above.

[0014] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to execute the aforementioned method.

[0015] According to a fifth aspect of the present disclosure, a computer program product is provided, comprising a computer program, which implements the aforementioned method when executed by a processor.

[0016] 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.

[0017] The solution provided in this embodiment can obtain a vehicle's detection results for a target obstacle in multiple dimensions. Based on these detection results and preset information about the target obstacle, a perception assessment result for the vehicle in each of these dimensions is obtained. In this way, by evaluating the vehicle's detection results for the obstacle in multiple dimensions, a perception assessment result for each dimension is obtained, ensuring efficient acquisition of the vehicle's perception assessment results while also ensuring accurate and comprehensive evaluation of the vehicle's detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings are provided to facilitate a better understanding of the present invention and do not constitute a limitation of the present disclosure.

[0019] Figure 1 is a flowchart of a perception evaluation method according to an embodiment of the present disclosure;

[0020] Figure 2 is a schematic diagram of a scenario of a perception evaluation method according to an embodiment of the present disclosure;

[0021] Figure 3 is another scenario schematic diagram of a perception assessment method according to an embodiment of the present disclosure;

[0022] Figure 4 is another scenario schematic diagram of a perception assessment method according to an embodiment of the present disclosure;

[0023] Figure 5 is a schematic diagram of the structure of a perception evaluation device according to another embodiment of the present disclosure;

[0024] Figure 6 It is a block diagram of an electronic device used to implement a perception evaluation method according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0025] 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.

[0026] The first embodiment of the present disclosure provides a perception evaluation method, such as Figure 1 Shown, including:

[0027] S101: Acquire detection results of a target obstacle in multiple dimensions by a vehicle; wherein the detection results in multiple dimensions are obtained by the vehicle during driving;

[0028] S102: Based on the detection results of the multiple dimensions and the preset information of the target obstacle, obtain a perception evaluation result of the vehicle in each of the multiple dimensions.

[0029] The perception assessment method provided in the first embodiment described above can be applied to electronic devices. Specifically, the electronic device can be a terminal device or a server, such as a laptop, tablet computer, desktop computer, ordinary server, cloud server, and the like. It should be understood that the above is merely an illustrative description of the execution entities capable of executing the perception assessment method provided in this embodiment. In actual processing, it is not limited to the devices mentioned in the above examples, as long as the examples are not exhaustive.

[0030] It can be seen that by adopting the above scheme, the vehicle's detection results of the target obstacle in multiple dimensions can be obtained; based on the detection results of the multiple dimensions and the preset information of the target obstacle, the perception evaluation results of the vehicle in each of the multiple dimensions can be obtained. In this way, by evaluating the vehicle's detection results of the obstacle in multiple dimensions, the vehicle's perception evaluation results in each dimension are obtained, thereby ensuring that the perception evaluation results of the vehicle are obtained efficiently while also ensuring accurate and comprehensive evaluation of the vehicle's detection results; and, through the above scheme, abnormal problems in the vehicle's perception of obstacles can be effectively and accurately located, thereby effectively promoting the optimization of the perception module algorithm, improving iteration efficiency, and reducing labor costs, thereby achieving the purpose of improving the safety and comfort of autonomous driving vehicles.

[0031] In some possible implementations, the aforementioned vehicle's detection results of the target obstacle in multiple dimensions may include at least one of the following: multiple detection frames of the target obstacle; the vehicle's attribute detection results of the target obstacle; and the vehicle's area detection frame of the target obstacle. Each of the multiple detection frames includes a detection result indicating whether the target obstacle is detected; and the target obstacle attribute detection results include one of the following: a crushable attribute, an uncrushable attribute, or a partially crushable attribute.

[0032] Accordingly, the preset information of the target obstacle may be pre-set according to the actual situation of the target obstacle. The preset information of the target obstacle is related to the detection results of the target obstacle by the vehicle in multiple dimensions.

[0033] Specifically, when the vehicle's multi-dimensional detection results of a target obstacle include multiple detection frames of the vehicle's target obstacle, the preset information of the target obstacle includes the position of the target obstacle, which specifically refers to the actual position of the target obstacle. The position of the target obstacle can be expressed in latitude and longitude.

[0034] In a case where the detection results of the vehicle on the target obstacle in multiple dimensions include the detection results of the vehicle on the attributes of the target obstacle, the preset information of the target obstacle includes the preset attributes of the target obstacle.

[0035] In a case where the vehicle's multi-dimensional detection results of a target obstacle include an area range detection frame of the vehicle for the target obstacle, the preset information of the target obstacle includes a safety detection frame and a near-host vehicle detection frame of the target obstacle; wherein the safety detection frame includes the area where the target obstacle is located; and the near-host vehicle detection frame includes the portion of the area where the target obstacle is located that is closest to the vehicle.

[0036] In one possible example, the vehicle's multi-dimensional detection results for a target obstacle include all of the following: multiple detection frames of the target obstacle, the vehicle's attribute detection results for the target obstacle, and the vehicle's area detection frame for the target obstacle. Accordingly, the preset information for the target obstacle includes all of the following: the target obstacle's location, its preset attributes, its safety detection frame, and its proximity detection frame.

[0037] In some possible implementations, obtaining the detection results of the vehicle for the target obstacle in multiple dimensions includes: obtaining multiple detection frames of the vehicle for the target obstacle; wherein each detection frame in the multiple detection frames includes a detection result indicating whether the target obstacle is detected;

[0038] The perception evaluation result of the vehicle in each of the multiple dimensions is obtained based on the detection results of the multiple dimensions and the preset information of the target obstacle, including: determining at least one of the farthest detection distance, the stable detection distance, the safe detection distance and the frame pass rate based on the multiple detection frames and the preset information of the target obstacle; and determining the perception evaluation result of the vehicle in the missed detection rate dimension based on at least one of the farthest detection distance, the stable detection distance, the safe detection distance and the frame pass rate.

[0039] The multiple detection frames of the target obstacle detected by the vehicle may be obtained from historical data stored by the vehicle.

[0040] The vehicle may obtain any one of the multiple detection frames for the target obstacle by: acquiring data from a collection unit in a current frame, inputting the data into a perception unit, and obtaining a detection frame output by the perception unit, wherein the detection frame includes a detection result of whether the target obstacle is detected. The collection unit specifically includes at least one of an image acquisition device, a point cloud acquisition device, and a posture-related sensor installed in the vehicle. The image acquisition device may include, but is not limited to, at least one of the following: a camera, a surround-view camera, an infrared camera, a camera, a time-of-flight (ToF) camera, etc.; the point cloud acquisition device may include, but is not limited to, a lidar and / or a radar; and the posture-related sensor may include, but is not limited to, at least one of the following: a global navigation satellite system (GNSS) and an inertial measurement unit (IMU). The perception unit may be a unit in the vehicle for processing the collected data, and may store a trained neural network model. Inputting the collected data into the perception unit to obtain the detection frame output by the perception unit specifically involves inputting the collected data into the neural network model in the perception unit to obtain the detection frame output by the neural network model in the perception unit.

[0041] By adopting the above solution, at least one of the maximum detection distance, stable detection distance, safe detection distance, and frame pass rate can be determined based on the multiple detection frames and the preset information of the target obstacle. Based on at least one of the maximum detection distance, the stable detection distance, the safe detection distance, and the frame pass rate, the perception evaluation result of the vehicle under the dimension of missed detection rate can be determined. In this way, the vehicle detection results can be classified, and the perception evaluation result of the vehicle under the dimension of missed detection rate can be obtained based on the multiple detection frames of the vehicle for the target obstacle and the preset information of the target obstacle in the vehicle detection results.

[0042] In some possible implementations, determining at least one of a maximum detection distance, a stable detection distance, a safe detection distance, and a frame pass rate based on the multiple detection frames and the preset information of the target obstacle includes at least one of the following:

[0043] Determining a first valid frame from one or more valid frames included in the multiple detection frames, and determining a first detected position of the vehicle based on a time corresponding to the first valid frame; obtaining the maximum detection distance based on the position of the target obstacle included in the preset information of the target obstacle and the first detected position of the vehicle; wherein each valid frame in the one or more valid frames is a detection frame in which a detection result indicates that the target obstacle is detected;

[0044] Determining a plurality of consecutive stable detection frames from the plurality of detection frames, and determining a stable detected position of the vehicle based on a time corresponding to a first detection frame among the plurality of consecutive stable detection frames; obtaining the stable detection distance based on the position of the target obstacle included in the preset information of the target obstacle and the stable detected position of the vehicle; wherein the number of consecutive missed detection frames included in the plurality of consecutive stable detection frames is less than a preset threshold value, and the missed detection frame is a detection frame in which a detection result indicates that the target obstacle was not detected;

[0045] Determining a last detection frame from the multiple detection frames based on a position of the target obstacle included in the preset information of the target obstacle; obtaining a total number of frames between the first valid frame and the last detection frame, and a number of valid frames between the first valid frame and the last detection frame, and obtaining a frame pass rate based on the number of valid frames and the total number of frames;

[0046] Determine a plurality of consecutive stable detection frames from the plurality of detection frames, determine a reaction distance of the vehicle based on the vehicle speed corresponding to the first stable detection frame in the plurality of consecutive stable detection frames, and obtain the safe detection distance based on the reaction distance of the vehicle and the braking distance of the vehicle.

[0047] The determining of the first valid frame from the one or more valid frames included in the multiple detection frames may be: determining one or more valid frames whose detection results are that the target obstacle is detected from the multiple detection frames, and taking the earliest valid frame among the one or more valid frames as the first valid frame. Figure 2 For example, in Figure 2 In the four scenes A to D, a white smiling face represents a valid frame. The direction of the arrow pointing from the vehicle to the obstacle in the four scenes A to D represents time from early to late. The first valid frame among the multiple valid frames in scene A is frame 201, the first valid frame in scene B is frame 211, the first valid frame in scene C is frame 221, and the first valid frame in scene D is frame 231.

[0048] Determining the first detected position of the vehicle based on the time corresponding to the first valid frame may include obtaining the time corresponding to the first valid frame, obtaining the vehicle's position information at the time corresponding to the first valid frame from the vehicle's historical data, and using the vehicle's position information as the first detected position of the vehicle. The vehicle's position information, i.e., the first detected position of the vehicle, may be expressed in latitude and longitude in an earth coordinate system.

[0049] The preset information of the obstacle may at least include the position of the target obstacle. The position of the target obstacle and the first detected position of the vehicle may both be represented by longitude and latitude.

[0050] Obtaining the farthest detection distance based on the position of the target obstacle included in the preset information of the target obstacle and the first detection position of the vehicle may refer to calculating the distance between the position of the target obstacle included in the preset information of the target obstacle and the first detection position of the vehicle, and using the distance as the farthest detection distance.

[0051] Determining a plurality of consecutive stable detection frames from the plurality of detection frames may include: determining, from the plurality of detection frames, one or more valid frames in which the target obstacle is detected, and one or more missed detection frames in which the target obstacle is not detected; arranging the one or more valid frames and the one or more missed detection frames in chronological order, and extracting consecutive valid frames in order from the latest valid frame to the earliest valid frame; and, if one or more missed detection frames exist after the plurality of consecutive valid frames, determining whether the number of the one or more missed detection frames is less than a preset threshold value; if so, extracting the one or more missed detection frames and continuing to extract valid frames; if not, stopping extracting the one or more missed detection frames and treating all currently extracted valid frames and all missed detection frames as the plurality of consecutive stable detection frames. The preset threshold value may be set according to actual circumstances, such as 4, 2, or a larger or smaller value, which is not exhaustive here.

[0052] Determining the stable detected position of the vehicle based on the time corresponding to the first detection frame among the plurality of consecutive stable detection frames can include: using the earliest detection frame among the plurality of consecutive stable detection frames as the first detection frame, obtaining the time corresponding to the first detection frame, obtaining the position information of the vehicle at the time corresponding to the first detection frame from the vehicle's historical data, and using the vehicle position information as the stable detected position of the vehicle. The stable detected position of the vehicle can be expressed using longitude and latitude in an earth coordinate system.

[0053] Combine Figure 2 For example, Figure 2 In the four scenes A to D, the black crying face represents the missed detection frame. For example, the preset threshold value is 4: Figure 2 The latest valid frame in scene A is Figure 2 Valid frame 202 in the valid frame 202, after extracting valid frames in the earlier direction from the valid frame 202, there are two consecutive missed frames (i.e. Figure 2 If the number of missed detection frames is less than 4, then continue to extract earlier valid frames until the valid frame 201 is extracted. There are 4 consecutive missed detection frames at a position earlier than the valid frame 201. The number of missed detection frames is not less than 4. Then, the continuous multiple stable detection frames in scene A are all detection frames between 201 and 202. Furthermore, the stable detection distance can be obtained based on the continuous multiple stable detection frames. Figure 2 The latest valid frame in scene B is Figure 2 Valid frame 212 in the valid frame 212, after extracting valid frames in the earlier direction, there are 4 consecutive missed frames (i.e. Figure 2The number of missed detection frames is greater than or equal to 4, and the continuous multiple stable detection frames in scene A are all detection frames between 213 and 212; further, a stable detection distance can be obtained based on the continuous multiple stable detection frames. Figure 2 The latest valid frame in scene C is Figure 2 Valid frame 222 is extracted from the valid frame 222 in an earlier direction, and no consecutive missed detection frames greater than or equal to 4 appear in the middle until the first valid frame 221 is extracted. Therefore, there are multiple consecutive stable detection frames from the valid frame 222 to the first valid frame 221. Further, a stable detection distance can be obtained based on the multiple consecutive stable detection frames. Figure 2 The latest valid frame in scene D is Figure 2 , valid frame 232 is continuously extracted from valid frame 232 toward an earlier direction, and no consecutive missed detection frames greater than or equal to 4 appear in the middle until valid frame 233 is extracted, while 4 consecutive missed detection frames (greater than or equal to 4) appear at a position earlier than valid frame 233. Therefore, there are multiple consecutive stable detection frames from valid frame 232 to valid frame 233. Further, a stable detection distance can be obtained based on the multiple consecutive stable detection frames.

[0054] Determining the last detection frame from the multiple detection frames based on the position of the target obstacle included in the preset information of the target obstacle may include: determining, based on the position of the target obstacle included in the preset information of the target obstacle, a time at which the position of the target obstacle coincides with the position of the vehicle; and determining the last detection frame among the multiple detection frames based on the time.

[0055] The total number of frames between the first valid frame and the last detected frame in the plurality of detected frames may be obtained by taking the number of all detected frames between the first valid frame and the last detected frame as the total number of frames. The specific description of the first valid frame is the same as that in the above embodiment and is not repeated here.

[0056] Obtaining the frame pass rate based on the number of valid frames and the total number of frames may be by calculating a ratio of the number of valid frames to the total number of frames, and using the ratio as the frame pass rate.

[0057] Determining the vehicle's reaction distance based on the vehicle's speed corresponding to the first stable detection frame among the plurality of consecutive stable detection frames may include: selecting the earliest stable detection frame from the plurality of consecutive stable detection frames as the first stable detection frame, determining the vehicle's speed at the time corresponding to the first stable detection frame; and multiplying the vehicle's speed at the time corresponding to the first stable detection frame by the single-frame time to obtain the vehicle's reaction distance. The vehicle's speed at the time corresponding to the first stable detection frame may be obtained from historical vehicle data. For example, the vehicle's reaction distance may be expressed using the following formula: reaction distance = single-frame time × vehicle speed at the time corresponding to the first stable detection frame. The single-frame time may refer to the vehicle's single-frame perception time, i.e., the time it takes for the vehicle's perception unit to obtain any detection frame. Specifically, it may be P99 (99th percentile), i.e., 99 frames per second.

[0058] The safe detection distance obtained based on the reaction distance of the vehicle and the braking distance of the vehicle may be obtained by adding the reaction distance of the vehicle and the braking distance of the vehicle to obtain a first result, and using the first result as the safe detection distance. The braking distance of the vehicle may specifically be obtained by obtaining the braking distance of the vehicle based on RSS (Responsibility Sensitive Safety); the braking distance obtained by RSS may be obtained based on the formula S=V*V / 2a; wherein S is the braking distance of the vehicle, V may be the speed of the vehicle at the moment corresponding to the first stable detection frame, and a is the deceleration of the vehicle. In a preferred embodiment, the deceleration of the vehicle does not exceed 5m / s2.

[0059] It should be noted that, based on the preset information of the multiple detection frames and the target obstacle, part or all of the maximum detection distance, stable detection distance, safe detection distance, and frame pass rate can be determined based on a preset strategy. The preset strategy can be set according to actual conditions. For example, if only the stable detection distance and the safe detection distance need to be analyzed this time, the preset strategy is set to determine the stable detection distance and the safe detection distance based on the multiple detection frames and the preset information of the target obstacle. For another example, if all of the maximum detection distance, stable detection distance, safe detection distance, and frame pass rate need to be analyzed this time, the preset strategy is set to determine the maximum detection distance, stable detection distance, safe detection distance, and frame pass rate based on the multiple detection frames and the preset information of the target obstacle.

[0060] It can be seen that by adopting the above scheme, at least one of the farthest detection distance, stable detection distance, safe detection distance and frame pass rate can be determined based on multiple detection frames of the vehicle to the target obstacle and the preset information of the target obstacle. In this way, multiple parameters under the missed detection rate dimension can be obtained through the detection frames and the preset information of the target obstacle to perform perception evaluation, and under the missed detection rate dimension, perception evaluation is also performed through the constraints of the four parameters of the farthest detection distance, stable detection distance, safe detection distance and frame pass rate, thereby ensuring the accuracy of the perception evaluation results.

[0061] In some possible embodiments, the determining of the perception evaluation result of the vehicle in the missed detection rate dimension based on at least one of the farthest detection distance, the stable detection distance, the safe detection distance, and the frame pass rate includes: when it is determined that a preset condition is met based on at least one of the farthest detection distance, the stable detection distance, the safe detection distance, and the frame pass rate, determining that the perception evaluation result of the vehicle in the missed detection rate dimension is abnormal; wherein the preset condition includes at least one of the following: the farthest detection distance is less than the preset detection distance, the stable detection distance is less than the preset detection distance, the safe detection distance is less than the preset safety distance, and the frame pass rate is less than the preset frame pass rate.

[0062] The preset detection distance can be set according to actual conditions, for example, it can be set to 10 meters, 20 meters, etc., which is not limited here.

[0063] The preset detection distance can be set according to actual conditions, for example, it can be smaller than the preset detection distance, and the preset detection distance can be 10 meters, 5 meters, etc.

[0064] The preset safety distance can be set according to actual conditions, for example, it can be 10 meters, or larger or smaller, and the list is not exhaustive.

[0065] The preset frame pass rate can be set according to actual conditions, for example, it can be 80%, 90% or larger or smaller, and the list is not exhaustive.

[0066] The above embodiments have explained that whether part or all of the maximum detection distance, the stable detection distance, the safe detection distance, and the frame pass rate are obtained this time is determined according to a preset strategy; accordingly, determining the perception evaluation result of the vehicle under the missed detection rate dimension based on part or all of the maximum detection distance, the stable detection distance, the safe detection distance, and the frame pass rate is also related to the preset strategy. For example, the preset strategy is set so that this processing needs to determine the stable detection distance and the safe detection distance based on the multiple detection frames and the preset information of the target obstacle; accordingly, the perception evaluation result of the vehicle under the missed detection rate dimension needs to be determined based on the stable detection distance and the safe detection distance. For another example, the preset strategy is set so that this processing needs to determine the maximum detection distance, the stable detection distance, the safe detection distance, and the frame pass rate based on the multiple detection frames and the preset information of the target obstacle; accordingly, the perception evaluation result of the vehicle under the missed detection rate dimension needs to be determined based on the maximum detection distance, the stable detection distance, the safe detection distance, and the frame pass rate.

[0067] Specifically, based on at least one of the farthest detection distance, the stable detection distance, the safe detection distance and the frame pass rate, the perception evaluation result of the vehicle under the missed detection rate dimension is determined, including: based on at least one of the farthest detection distance, the stable detection distance, the safe detection distance and the frame pass rate, judging whether a preset condition is met; if the preset condition is met, determining that the perception evaluation result of the vehicle under the missed detection rate dimension is abnormal; if the preset condition is not met, determining that the perception evaluation result of the vehicle under the missed detection rate dimension is normal.

[0068] The aforementioned satisfaction of the preset condition may refer to: when any one of the following conditions is met: the farthest detection distance is less than the preset detection distance, the stable detection distance is less than the preset detection distance, the safe detection distance is less than the preset safety distance, and the frame pass rate is less than the preset frame pass rate, it is determined that the preset condition is satisfied.

[0069] The aforementioned failure to meet the preset conditions may refer to: when the farthest detection distance is not less than the preset detection distance, the stable detection distance is not less than the preset detection distance, the safe detection distance is not less than the preset safety distance, and the frame pass rate is not less than the preset frame pass rate, it is determined that the preset conditions are not met.

[0070] For example, the preset strategy is set so that this processing needs to determine the stable detection distance and the safe detection distance based on the multiple detection frames and the preset information of the target obstacle; accordingly, this time it is necessary to judge whether the stable detection distance is less than the preset detection distance and whether the safe detection distance is less than the preset safety distance. If the stable detection distance is less than the preset detection distance, and / or the safe detection distance is less than the preset safety distance, then it is determined that the perception evaluation result of the vehicle under the missed detection rate dimension is abnormal; if the stable detection distance is not less than the preset detection distance, and the safe detection distance is not less than the preset safety distance, then it is determined that the perception evaluation result of the vehicle under the missed detection rate dimension is normal.

[0071] In some possible examples, the method may further include: when the stable detection distance is less than or equal to the safe detection distance, determining that the perception evaluation result of the vehicle under the missed detection rate dimension is abnormal.

[0072] As can be seen, by adopting the above solution, it is possible to determine whether the vehicle's perception assessment result under the missed detection rate dimension is abnormal based on at least one of the maximum detection distance, the stable detection distance, the safe detection distance, and the frame pass rate. In this way, multiple parameters can be used under the missed detection rate dimension to determine whether the perception assessment result is abnormal, thereby making the assessment result more accurate.

[0073] In some possible implementations, obtaining the vehicle's detection results of a target obstacle in multiple dimensions includes obtaining the vehicle's attribute detection results of the target obstacle. Obtaining a perception evaluation result of the vehicle in each of the multiple dimensions based on the detection results in the multiple dimensions and preset information about the target obstacle includes determining a state of overlap between the vehicle and the target obstacle based on the vehicle's stop position and a detection frame of the target obstacle in the preset information about the target obstacle; and determining a perception evaluation result of the vehicle in the attribute dimension based on the state of overlap between the vehicle and the target obstacle, the attribute detection results, and preset attributes of the target obstacle in the preset information about the target obstacle.

[0074] The attribute detection result of the target obstacle includes one of the following: a crushable attribute, an uncrushable attribute, and a partially crushable attribute.

[0075] The determining of the overlap state between the vehicle and the target obstacle based on the vehicle's stopping position and the detection frame of the target obstacle in the preset information of the target obstacle may be: determining the overlap state between the vehicle and the target obstacle based on whether an area where the vehicle's stopping position is located overlaps with an area where the detection frame of the target obstacle in the preset information of the target obstacle is located.

[0076] Determining the perception evaluation result of the vehicle in the attribute dimension based on the overlap state between the vehicle and the target obstacle, the attribute detection result, and the preset attributes of the target obstacle in the preset information of the target obstacle may include: determining the perception evaluation result of the vehicle in the attribute dimension based on the attribute detection result and the preset attributes of the target obstacle in the preset information of the target obstacle when the overlap state between the vehicle and the target obstacle is non-overlap; and determining the perception evaluation result of the vehicle in the attribute dimension based on the attribute detection result and the preset attributes of the target obstacle in the preset information of the target obstacle when the overlap state between the vehicle and the target obstacle is overlap. The overlap state between the vehicle and the target obstacle indicates that the vehicle and the target obstacle have collided.

[0077] It can be seen that by adopting the above scheme, the perception evaluation result of the vehicle in the attribute dimension can be determined through the overlap state between the vehicle and the target obstacle, as well as the attribute detection result and the preset attributes of the target obstacle in the preset information of the target obstacle, so that the attribute perception dimension of the vehicle can be evaluated.

[0078] In some possible implementations, determining a perception evaluation result of the vehicle in an attribute dimension based on a state of overlap between the vehicle and the target obstacle, a result of the vehicle detecting an attribute of the target obstacle, and a preset attribute of the target obstacle in the preset information of the target obstacle includes one of the following:

[0079] When the overlap state between the vehicle and the target obstacle is non-overlap, and the attribute detection result is different from the preset attribute of the target obstacle in the preset information of the target obstacle, determining that the perception evaluation result of the vehicle in the attribute dimension is abnormal;

[0080] When the overlap state between the vehicle and the target obstacle is overlap, the attribute detection result is a crushable attribute or a partially crushable attribute, and the preset attribute of the target obstacle in the preset information of the target obstacle is an uncrushable attribute, determining that the perception evaluation result of the vehicle in the attribute dimension is abnormal;

[0081] When the overlapping state of the vehicle and the target obstacle is overlapping, the attribute detection result is a partially crushable attribute, and the preset attribute of the target obstacle in the preset information of the target obstacle is a partially crushable attribute, the chassis height of the vehicle and the height of the target obstacle in the preset information of the target obstacle are obtained. When the chassis height of the vehicle is less than the height of the target obstacle, it is determined that the perception evaluation result of the vehicle in the attribute dimension is abnormal.

[0082] When the overlap state between the vehicle and the target obstacle is non-overlap, and the attribute detection result is different from the preset attribute of the target obstacle in the preset information of the target obstacle, determining that the perception evaluation result of the vehicle in the attribute dimension is abnormal may include one of the following:

[0083] When the overlap state between the vehicle and the target obstacle is non-overlap, if the attribute detection result is crushable, and the preset attribute of the target obstacle in the preset information of the target obstacle is non-crushable or partially crushable, then determining that the perception evaluation result of the vehicle in the attribute dimension is abnormal;

[0084] When the overlap state between the vehicle and the target obstacle is non-overlap, if the attribute detection result is partially crushable, and the preset attribute of the target obstacle in the preset information of the target obstacle is non-crushable or crushable, then determining that the perception evaluation result of the vehicle in the attribute dimension is abnormal;

[0085] When the overlap state between the vehicle and the target obstacle is non-overlap, if the attribute detection result is that the vehicle cannot be crushed, and the preset attribute of the target obstacle in the preset information of the target obstacle is that the vehicle can be crushed or partially crushed, then the perception evaluation result of the vehicle under the attribute dimension is determined to be abnormal.

[0086] The vehicle's chassis height can be obtained directly from the vehicle's attribute information. The vehicle's attribute information can be pre-set, for example, before the vehicle is evaluated, the vehicle's attribute information can be pre-set, and the vehicle's attribute information can include at least the vehicle's chassis height. Figure 3 For example, the chassis height of the vehicle is Figure 3 S301 is the chassis height of the vehicle, which is specifically the height between the chassis of the vehicle and the ground; the height of the target obstacle in the preset information of the target obstacle is Figure 3 The height between the highest point of the obstacle and the ground in S302 is shown. Figure 3It can be seen that if the target obstacle height S302 is greater than the vehicle's chassis height S301, even if the vehicle's attribute detection result indicates that the target obstacle is partially crushable, but the target obstacle height S302 is greater than the vehicle's chassis height S301, the vehicle's attribute detection result can still be determined to be abnormal. In other words, if the vehicle and the target obstacle overlap, the vehicle's chassis height and the target obstacle height are further used to determine whether the vehicle has collided, thereby determining the vehicle's perception assessment result under the attribute dimension.

[0087] In addition, determining the perception evaluation result of the vehicle in the attribute dimension based on the overlap state of the vehicle and the target obstacle, the attribute detection result of the vehicle for the target obstacle, and the preset attribute of the target obstacle in the preset information of the target obstacle may also include at least one of the following: determining that the perception evaluation result of the vehicle in the attribute dimension is normal when the overlap state of the vehicle and the target obstacle is non-overlap and the attribute detection result is the same as the preset attribute of the target obstacle in the preset information of the target obstacle; and determining that the perception evaluation result of the vehicle in the attribute dimension is normal when the overlap state of the vehicle and the target obstacle is overlap, the attribute detection result is a crushable attribute, and the preset attribute of the target obstacle in the preset information of the target obstacle is a crushable attribute.

[0088] By adopting the above solution, the vehicle's perception assessment result under the attribute dimension can be determined to be abnormal based on the overlap between the vehicle and the target obstacle, the vehicle's attribute detection results for the target obstacle, and the preset attributes of the target obstacle in the preset information of the target obstacle. In this way, the vehicle's perception assessment result under the attribute dimension can be determined to be abnormal, ensuring the comprehensiveness and accuracy of the final assessment result. In particular, under the partially crushable attribute, even if the vehicle's detection result for the target obstacle is the same as the preset attributes of the target obstacle, the vehicle's chassis height and the height of the target obstacle are further used to determine whether the vehicle has collided, thereby determining the vehicle's perception assessment result under the attribute dimension, thus ensuring the accuracy of the final assessment result.

[0089] In some possible implementations, obtaining the vehicle's detection results of a target obstacle in multiple dimensions includes obtaining a regional range detection frame for the target obstacle. Obtaining a perception assessment result of the vehicle in each of the multiple dimensions based on the detection results in the multiple dimensions and preset information about the target obstacle includes determining the vehicle's perception assessment result in the regional range dimension based on the regional range detection frame, a safety detection frame in the preset information about the target obstacle, and a near-host vehicle detection frame; wherein the safety detection frame includes the area where the target obstacle is located, and the near-host vehicle detection frame includes the portion of the target obstacle's area closest to the vehicle.

[0090] Combine Figure 4 , the area range detection frame, the safety detection frame, and the near-host vehicle detection frame are exemplarily described, Figure 4 In the figure, the actual position of the target obstacle is represented by a dotted-line frame 400, the area detection frame of the vehicle to the target obstacle is represented by a dotted-line frame 401, the safety detection frame in the preset information of the target obstacle is represented by a solid-line frame 411, and the near-host vehicle detection frame in the preset information of the target obstacle is represented by a solid-line frame 412. Figure 4 The shapes, sizes, and positions of the above detection frames are only exemplary representations, and the specific shapes, sizes, and positions of the detection frames depend on the actual situation.

[0091] pass Figure 4 It can be seen that there may be some differences between the area range detection frame detected by the vehicle and the actual safety detection frame and near-host vehicle detection frame of the target obstacle. By adopting the solution provided by this embodiment, the perception assessment result of the vehicle in the area range dimension can be determined based on these three detection frames. In some possible examples, the perception assessment result of the vehicle in the area range dimension can be determined based on the area range detection frame, the safety detection frame in the preset information of the target obstacle, and the near-host vehicle detection frame. This can be done by calculating a first overlap ratio between the area range detection frame and the safety detection frame in the preset information of the target obstacle, and calculating a second overlap ratio between the area range detection frame and the near-host vehicle detection frame in the preset information of the target obstacle. If the first overlap ratio is less than a first ratio threshold and the second overlap ratio is less than a second ratio threshold, the perception assessment result of the vehicle in the area range dimension is determined to be abnormal. It can also include determining that the perception assessment result of the vehicle in the area range dimension is normal if the first overlap ratio is not less than the first ratio threshold and the second overlap ratio is not less than the second ratio threshold. The first ratio threshold and the second ratio threshold can be set according to actual conditions and are not limited here.

[0092] By adopting the above solution, the perception evaluation result of the vehicle in the area range dimension can be determined through the area range detection frame, the safety detection frame in the preset information of the target obstacle, and the near-host vehicle detection frame, so that the evaluation result can be more accurate and comprehensive.

[0093] In some possible embodiments, determining the perception evaluation result of the vehicle in the area range dimension based on the area range detection frame, the safety detection frame in the preset information of the target obstacle, and the near-host vehicle detection frame includes: determining a first intersection and a first union between the area range detection frame and the safety detection frame in the preset information of the target obstacle, and determining a first ratio based on the first intersection and the first union; determining a second intersection between the area range detection frame and the near-host vehicle detection frame in the preset information of the target obstacle, and determining a second ratio based on the second intersection and the near-host vehicle detection frame; when the first ratio is less than a first threshold value, and / or the second ratio is less than a second threshold value, determining that the perception evaluation result of the vehicle in the area range dimension is abnormal.

[0094] Determining a first intersection and a first union between the area detection frame and the safety detection frame in the preset information of the target obstacle, and determining the first ratio based on the first intersection and the first union may include calculating a first intersection between the area detection frame and the safety detection frame in the preset information of the target obstacle, calculating a first union between the area detection frame and the safety detection frame in the preset information of the target obstacle, and dividing the first intersection by the first union to obtain the first ratio. The first ratio may also be referred to as an average intersection over union (IoU).

[0095] Calculating the first intersection between the area detection frame and the safety detection frame in the preset information of the target obstacle may specifically refer to calculating a first overlapping area between the area detection frame and the safety detection frame in the preset information of the target obstacle, and using the first overlapping area as the first intersection. Calculating the first union between the area detection frame and the safety detection frame in the preset information of the target obstacle may refer to calculating a first overlapping area between the area detection frame and the safety detection frame in the preset information of the target obstacle, a first remaining area of ​​the area detection frame excluding the overlapping area, and a second remaining area of ​​the safety detection frame excluding the overlapping area, and using the sum of the first overlapping area, the first remaining area, and the second remaining area as the first union.

[0096] Assuming that the area detection box is represented as "pred_polygon" and the safety detection box in the preset information of the target obstacle is represented as "check_polygon", the calculation method of the first ratio can be expressed by the following formula:

[0097] IoU = Intersection(pred_polygon, check_polygon) / Union(pred_polygon, check_polygon). IoU is the first ratio, intersection(*) means calculating the intersection, and Union(*) means calculating the union.

[0098] Determining a second intersection between the area detection frame and the near-host vehicle detection frame in the preset information about the target obstacle, and determining a second ratio based on the second intersection and the near-host vehicle detection frame may include calculating a second intersection between the area detection frame and the near-host vehicle detection frame in the preset information about the target obstacle, and dividing the second intersection by the near-host vehicle detection frame to obtain the second ratio. The second ratio may also be referred to as an average intersection ratio (IoN).

[0099] Calculating a second intersection between the area detection frame and the near-host vehicle detection frame in the preset information about the target obstacle may include calculating a second overlapping area between the area detection frame and the near-host vehicle detection frame in the preset information about the target obstacle, and using the second overlapping area as the second intersection. Dividing the second intersection by the near-host vehicle detection frame to obtain a second ratio may include dividing the second intersection by the area of ​​the near-host vehicle detection frame to obtain the second ratio.

[0100] Assuming that the area detection box is still represented as "pred_polygon", the detection box closest to the main vehicle in the preset information of the target obstacle is represented as "nearest_gt_polygon", where gt is the groundtruth, that is, the correctly labeled data. The calculation method of the second ratio can be expressed as follows:

[0101] IoN=Intersection(pred_polygon, nearest_gt_polygon) / nearest_gt_polygon; wherein IoN is the second ratio, and intersection(*) indicates calculating the intersection.

[0102] In addition, the aforementioned method may further include: when the first ratio is not less than a first threshold value and the second ratio is not less than a second threshold value, determining that the perception evaluation result of the vehicle in the area range dimension is normal.

[0103] The above-mentioned regional range detection frame of the vehicle for the target obstacle uses a safety detection frame and a near-host vehicle detection frame to obtain two ratios respectively, and then determines whether the perception evaluation result of the vehicle in the regional range dimension is normal based on the thresholds corresponding to the two ratios. This processing method is particularly suitable for scenarios such as temporary construction areas or temporary accident areas on the road. In this scenario, the temporary construction area or temporary accident area is the target obstacle in this embodiment. In this scenario, the vehicle's regional range detection frame for the target obstacle often has two problems. One is "creating something out of nothing", that is, mistakenly aggregating areas without construction elements; the other is excessive aggregation, that is, the construction boundary is associated with fence heads, columns, etc. By adopting the above-mentioned solution provided by this embodiment, the vehicle can accurately and efficiently judge the regional range detection frame of the target obstacle.

[0104] It should be understood that the above is only an exemplary explanation of the target obstacle being a temporary construction area or a temporary accident area on the road. In actual processing, the target obstacle in this embodiment is not limited to the above one, but can also be a pedestrian, an empty manhole cover, etc., which are not enumerated here.

[0105] By adopting the above scheme, the abnormality of the perception evaluation result of the vehicle in the area range dimension can be determined by the ratio of the intersection and / or union between the area range detection frame, the safety detection frame in the preset information of the target obstacle, and the near-host vehicle detection frame. For example, in the case where the target obstacle is a construction / accident area, two ratios are obtained based on the safety detection frame and the near-host vehicle detection frame of the vehicle's area range detection frame of the target obstacle, so that it can be accurately judged whether the vehicle's area range detection frame of the target obstacle aggregates areas without construction elements, and / or the aggregation is too large (for example, construction boundaries are associated with fence heads, columns, etc.), thereby ensuring a more accurate and comprehensive perception evaluation of the vehicle's area range detection frame of the target obstacle, providing a comprehensive reference for subsequent perception algorithm optimization, and thus ensuring the safety of the vehicle's autonomous driving.

[0106] It should be noted that the perception evaluation results of the aforementioned vehicle in each of the multiple dimensions can be used in combination in their entirety, or at least partially.

[0107] For example, multiple detection frames of the vehicle for the target obstacle can be obtained, and the attribute detection results of the vehicle for the target obstacle can be obtained; the perception evaluation results of the vehicle in the missed detection rate dimension and the perception evaluation results of the vehicle in the attribute dimension can be obtained.

[0108] For example, multiple detection frames of the vehicle for the target obstacle, the attribute detection results of the vehicle for the target obstacle, and the area range detection frame of the vehicle for the target obstacle can be obtained; the perception evaluation results of the vehicle in the missed detection rate dimension, the perception evaluation results of the vehicle in the attribute dimension, and the perception evaluation results of the vehicle in the area range dimension can be obtained.

[0109] It should also be noted that, after the processing provided by the aforementioned embodiments, if any one of the vehicle's perception evaluation results in the missed detection rate dimension, the vehicle's perception evaluation results in the attribute dimension, or the vehicle's perception evaluation results in the area range dimension is determined to be abnormal, it can be determined that the vehicle's perception calculation or perception unit requires optimization in the corresponding dimension. For example, if the vehicle's perception evaluation results in the area range dimension and the vehicle's perception evaluation results in the attribute dimension are determined to be abnormal, it can be determined that the vehicle's perception calculation or perception unit requires optimization in both dimensions.

[0110] Furthermore, after determining that the vehicle's perception calculation or perception unit needs to be optimized in one or more dimensions, an optimization notification can be generated and uploaded to the cloud. The cloud will then assign optimization tasks to the processing personnel corresponding to each of the one or more dimensions. The corresponding processing personnel will then iteratively update the algorithm (such as a neural network model) used by the vehicle's perception calculation or perception unit, and re-distribute the updated algorithm (such as a neural network model) to the vehicle. Through the above solution, the vehicle's autonomous driving perception detection results can be accurately measured, abnormal problems can be located, and based on the characteristics of the abnormal problems and the safety hazards they involve, an effective optimization direction for the perception module algorithm can be formulated, ensuring that the perception module algorithm is more accurate, improving iteration efficiency, and reducing labor costs, thereby achieving the goal of improving the safety and comfort of autonomous driving vehicles.

[0111] A second aspect of the present disclosure provides a perception evaluation device, such as Figure 5 Shown, including:

[0112] An acquisition module 501 is configured to acquire detection results of a target obstacle in multiple dimensions by a vehicle; wherein the detection results in multiple dimensions are obtained by the vehicle during driving;

[0113] The evaluation module 502 is configured to obtain a perception evaluation result of the vehicle in each of the multiple dimensions based on the detection results in the multiple dimensions and preset information of the target obstacle.

[0114] The acquisition module is configured to acquire a plurality of detection frames of the vehicle detecting the target obstacle; wherein each of the plurality of detection frames includes a detection result indicating whether the target obstacle is detected;

[0115] The evaluation module is used to determine at least one of the farthest detection distance, the stable detection distance, the safe detection distance and the frame pass rate based on the multiple detection frames and the preset information of the target obstacle; and determine the perception evaluation result of the vehicle under the missed detection rate dimension based on at least one of the farthest detection distance, the stable detection distance, the safe detection distance and the frame pass rate.

[0116] The evaluation module is configured to perform at least one of the following:

[0117] Determining a first valid frame from one or more valid frames included in the multiple detection frames, and determining a first detected position of the vehicle based on a time corresponding to the first valid frame; obtaining the maximum detection distance based on the position of the target obstacle included in the preset information of the target obstacle and the first detected position of the vehicle; wherein each valid frame in the one or more valid frames is a detection frame in which a detection result indicates that the target obstacle is detected;

[0118] Determining a plurality of consecutive stable detection frames from the plurality of detection frames, and determining a stable detected position of the vehicle based on a time corresponding to a first detection frame among the plurality of consecutive stable detection frames; obtaining the stable detection distance based on the position of the target obstacle included in the preset information of the target obstacle and the stable detected position of the vehicle; wherein the number of consecutive missed detection frames included in the plurality of consecutive stable detection frames is less than a preset threshold value, and the missed detection frame is a detection frame in which a detection result indicates that the target obstacle was not detected;

[0119] Determining a last detection frame from the multiple detection frames based on a position of the target obstacle included in the preset information of the target obstacle; obtaining a total number of frames between the first valid frame and the last detection frame, and a number of valid frames between the first valid frame and the last detection frame, and obtaining a frame pass rate based on the number of valid frames and the total number of frames;

[0120] Determine a plurality of consecutive stable detection frames from the plurality of detection frames, determine a reaction distance of the vehicle based on the vehicle speed corresponding to the first stable detection frame in the plurality of consecutive stable detection frames, and obtain the safe detection distance based on the reaction distance of the vehicle and the braking distance of the vehicle.

[0121] The evaluation module is configured to determine that a perception evaluation result of the vehicle under the missed detection rate dimension is abnormal when a preset condition is determined to be satisfied based on at least one of the maximum detection distance, the stable detection distance, the safe detection distance, and the frame pass rate;

[0122] Among them, the preset conditions include at least one of the following: the farthest detection distance is less than the preset detection distance, the stable detection distance is less than the preset detection distance, the safe detection distance is less than the preset safety distance, and the frame pass rate is less than the preset frame pass rate.

[0123] The acquisition module is used to obtain the attribute detection result of the vehicle on the target obstacle;

[0124] The evaluation module is configured to determine a state of overlap between the vehicle and the target obstacle based on the vehicle's stopped position and a detection frame of the target obstacle in the preset information of the target obstacle; and to determine a perception evaluation result of the vehicle in an attribute dimension based on the state of overlap between the vehicle and the target obstacle, the attribute detection result, and the preset attributes of the target obstacle in the preset information of the target obstacle.

[0125] The evaluation module is configured to perform one of the following:

[0126] When the overlap state between the vehicle and the target obstacle is non-overlap, and the attribute detection result is different from the preset attribute of the target obstacle in the preset information of the target obstacle, determining that the perception evaluation result of the vehicle in the attribute dimension is abnormal;

[0127] When the overlap state between the vehicle and the target obstacle is overlap, the attribute detection result is a crushable attribute or a partially crushable attribute, and the preset attribute of the target obstacle in the preset information of the target obstacle is an uncrushable attribute, determining that the perception evaluation result of the vehicle in the attribute dimension is abnormal;

[0128] When the overlapping state of the vehicle and the target obstacle is overlapping, the attribute detection result is a partially crushable attribute, and the preset attribute of the target obstacle in the preset information of the target obstacle is a partially crushable attribute, the chassis height of the vehicle and the height of the target obstacle in the preset information of the target obstacle are obtained. When the chassis height of the vehicle is less than the height of the target obstacle, it is determined that the perception evaluation result of the vehicle in the attribute dimension is abnormal.

[0129] The acquisition module is used to obtain the area range detection frame of the vehicle to the target obstacle;

[0130] The evaluation module is used to determine the perception evaluation result of the vehicle in the area range dimension based on the area range detection frame, the safety detection frame in the preset information of the target obstacle, and the near-host vehicle detection frame; wherein, the safety detection frame includes the area where the target obstacle is located; the near-host vehicle detection frame includes the portion of the area where the target obstacle is located that is closest to the vehicle.

[0131] The evaluation module is used to determine a first intersection and a first union between the area range detection frame and the safety detection frame in the preset information of the target obstacle, and determine a first ratio based on the first intersection and the first union; determine a second intersection between the area range detection frame and the near-host vehicle detection frame in the preset information of the target obstacle, and determine a second ratio based on the second intersection and the near-host vehicle detection frame; when the first ratio is less than a first threshold and / or the second ratio is less than a second threshold, determine that the perception evaluation result of the vehicle in the area range dimension is abnormal.

[0132] The perception evaluation device provided in this embodiment can be provided in an electronic device. The specific processing of each module in the device of this embodiment is the same as that in the aforementioned perception evaluation method, and will not be repeated here.

[0133] In the technical solutions disclosed herein, the acquisition, storage, and application of user personal information involved comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0134] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a vehicle, a readable storage medium, and a computer program product.

[0135] Figure 6A schematic block diagram of an example electronic device 600 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.

[0136] like Figure 6 As shown, the electronic device 600 includes a computing unit 601, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 into a random access memory (RAM) 603. Various programs and data required for the operation of the electronic device 600 can also be stored in the RAM 603. The computing unit 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0137] Multiple components in the electronic device 600 are connected to the I / O interface 605, including an input unit 606, such as a keyboard, a mouse, etc.; an output unit 607, such as various types of displays, speakers, etc.; a storage unit 608, such as a magnetic disk, an optical disk, etc.; and a communication unit 609, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 609 allows the electronic device 600 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0138] The computing unit 601 can be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 601 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, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 601 performs the various methods and processes described above. For example, in some embodiments, the various methods described above can be implemented as a computer software program that is tangibly contained in a machine-readable medium, such as a storage unit 608. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 600 via the ROM 602 and / or the communication unit 609. When the computer program is loaded into the RAM 603 and executed by the computing unit 601, one or more steps of the various methods described above can be performed. Alternatively, in other embodiments, the computing unit 601 can be configured to perform the various methods described above by any other appropriate means (e.g., by means of firmware).

[0139] Various embodiments of the systems and techniques described herein 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), 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.

[0140] 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.

[0141] 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.

[0142] 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).

[0143] 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.

[0144] 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.

[0145] 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 limited herein.

[0146] 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 perception assessment method, comprising: Obtaining detection results of the vehicle on a target obstacle in multiple dimensions; wherein the detection results in multiple dimensions are obtained by the vehicle during driving; Obtaining a perception evaluation result of the vehicle in each of the multiple dimensions based on the detection results of the multiple dimensions and preset information of the target obstacle; The acquiring of detection results of the target obstacle by the vehicle in multiple dimensions includes: acquiring attribute detection results of the target obstacle by the vehicle; obtaining a perception evaluation result of the vehicle in each of the multiple dimensions based on the detection results in the multiple dimensions and preset information of the target obstacle, including: determining an overlap state between the vehicle and the target obstacle based on the vehicle's stop position and a detection frame of the target obstacle in the preset information of the target obstacle; and determining a perception evaluation result of the vehicle in the attribute dimension based on the overlap state between the vehicle and the target obstacle, the attribute detection result, and preset attributes of the target obstacle in the preset information of the target obstacle. Wherein, the determining of the perception evaluation result of the vehicle in the attribute dimension based on the overlap state of the vehicle and the target obstacle, the attribute detection result of the vehicle for the target obstacle, and the preset attribute of the target obstacle in the preset information of the target obstacle includes: determining that the perception evaluation result of the vehicle in the attribute dimension is abnormal when the overlap state of the vehicle and the target obstacle is non-overlapping and the attribute detection result is different from the preset attribute of the target obstacle in the preset information of the target obstacle; determining that the perception evaluation result of the vehicle in the attribute dimension is abnormal when the overlap state of the vehicle and the target obstacle is overlapping and the attribute detection result is a crushable attribute or a partially crushable attribute. , and the preset attribute of the target obstacle in the preset information of the target obstacle is the non-crushable attribute, the perception evaluation result of the vehicle in the attribute dimension is determined to be abnormal; when the overlap state of the vehicle and the target obstacle is overlap, the attribute detection result is the partially crushable attribute, and the preset attribute of the target obstacle in the preset information of the target obstacle is the partially crushable attribute, the chassis height of the vehicle and the height of the target obstacle in the preset information of the target obstacle are obtained, and when the chassis height of the vehicle is less than the height of the target obstacle, the perception evaluation result of the vehicle in the attribute dimension is determined to be abnormal.

2. The method according to claim 1, wherein The obtaining of detection results of the target obstacle by the vehicle in multiple dimensions includes: obtaining multiple detection frames of the target obstacle by the vehicle; wherein each detection frame in the multiple detection frames includes a detection result indicating whether the target obstacle is detected; The perception evaluation result of the vehicle in each of the multiple dimensions is obtained based on the detection results of the multiple dimensions and the preset information of the target obstacle, including: determining at least one of the farthest detection distance, the stable detection distance, the safe detection distance and the frame pass rate based on the multiple detection frames and the preset information of the target obstacle; and determining the perception evaluation result of the vehicle in the missed detection rate dimension based on at least one of the farthest detection distance, the stable detection distance, the safe detection distance and the frame pass rate.

3. The method according to claim 2, wherein: The determining, based on the plurality of detection frames and the preset information of the target obstacle, at least one of the farthest detection distance, the stable detection distance, the safe detection distance, and the frame pass rate comprises at least one of the following: Determining a first valid frame from one or more valid frames included in the multiple detection frames, and determining a first detected position of the vehicle based on a time corresponding to the first valid frame; obtaining the maximum detection distance based on the position of the target obstacle included in the preset information of the target obstacle and the first detected position of the vehicle; wherein each valid frame in the one or more valid frames is a detection frame in which a detection result indicates that the target obstacle is detected; Determining a plurality of consecutive stable detection frames from the plurality of detection frames, and determining a stable detected position of the vehicle based on a time corresponding to a first detection frame among the plurality of consecutive stable detection frames; obtaining the stable detection distance based on the position of the target obstacle included in the preset information of the target obstacle and the stable detected position of the vehicle; wherein the number of consecutive missed detection frames included in the plurality of consecutive stable detection frames is less than a preset threshold value, and the missed detection frame is a detection frame in which a detection result indicates that the target obstacle was not detected; Determining a last detection frame from the multiple detection frames based on a position of the target obstacle included in the preset information of the target obstacle; obtaining a total number of frames between the first valid frame and the last detection frame, and a number of valid frames between the first valid frame and the last detection frame, and obtaining a frame pass rate based on the number of valid frames and the total number of frames; Determine a plurality of consecutive stable detection frames from the plurality of detection frames, determine a reaction distance of the vehicle based on the vehicle speed corresponding to the first stable detection frame in the plurality of consecutive stable detection frames, and obtain the safe detection distance based on the reaction distance of the vehicle and the braking distance of the vehicle.

4. The method according to claim 2, wherein: The determining, based on at least one of the farthest detection distance, the stable detection distance, the safe detection distance, and the frame pass rate, a perception evaluation result of the vehicle in the missed detection rate dimension includes: If it is determined that a preset condition is satisfied based on at least one of the maximum detection distance, the stable detection distance, the safe detection distance, and the frame pass rate, determining that a perception evaluation result of the vehicle under the missed detection rate dimension is abnormal; Among them, the preset conditions include at least one of the following: the farthest detection distance is less than the preset detection distance, the stable detection distance is less than the preset detection distance, the safe detection distance is less than the preset safety distance, and the frame pass rate is less than the preset frame pass rate.

5. The method according to any one of claims 1 to 4, wherein: The obtaining of detection results of the vehicle on the target obstacle in multiple dimensions includes: obtaining a region range detection frame of the vehicle on the target obstacle; The perception evaluation result of the vehicle in each of the multiple dimensions is obtained based on the detection results of the multiple dimensions and the preset information of the target obstacle, including: determining the perception evaluation result of the vehicle in the regional range dimension based on the regional range detection frame, the safety detection frame in the preset information of the target obstacle, and the near-host vehicle detection frame; wherein the safety detection frame includes the area where the target obstacle is located; the near-host vehicle detection frame includes the portion of the area where the target obstacle is located that is closest to the vehicle.

6. The method according to claim 5, wherein: The determining of a perception evaluation result of the vehicle in an area range dimension based on the area range detection frame, the safety detection frame in the preset information of the target obstacle, and the near-host vehicle detection frame includes: Determining a first intersection and a first union between the area range detection frame and the safety detection frame in the preset information of the target obstacle, and determining a first ratio based on the first intersection and the first union; determining a second intersection between the area range detection frame and the near-host vehicle detection frame in the preset information of the target obstacle, and determining a second ratio based on the second intersection and the near-host vehicle detection frame; When the first ratio is smaller than a first threshold value, and / or the second ratio is smaller than a second threshold value, it is determined that the perception evaluation result of the vehicle in the area range dimension is abnormal.

7. A perception assessment device, comprising: An acquisition module is used to obtain the vehicle's detection results of multiple dimensions of the target obstacle; wherein the detection results of the multiple dimensions are obtained by the vehicle during driving; an evaluation module, configured to obtain a perception evaluation result of the vehicle in each of the multiple dimensions based on the detection results of the multiple dimensions and preset information of the target obstacle; The acquisition module is configured to acquire an attribute detection result of the target obstacle by the vehicle; the evaluation module is configured to determine an overlap state between the vehicle and the target obstacle based on the vehicle's stop position and a detection frame of the target obstacle in the preset information of the target obstacle; and determine a perception evaluation result of the vehicle in the attribute dimension based on the overlap state between the vehicle and the target obstacle, the attribute detection result, and the preset attributes of the target obstacle in the preset information of the target obstacle. The evaluation module is configured to perform the following processing: determining that a perception evaluation result of the vehicle in the attribute dimension is abnormal if the overlap state between the vehicle and the target obstacle is non-overlap and the attribute detection result is different from a preset attribute of the target obstacle in the preset information of the target obstacle; determining that the perception evaluation result of the vehicle in the attribute dimension is abnormal if the overlap state between the vehicle and the target obstacle is overlap, the attribute detection result is a crushable attribute or a partially crushable attribute, and the preset attribute of the target obstacle in the preset information of the target obstacle is an uncrushable attribute; and obtaining a chassis height of the vehicle and a height of the target obstacle in the preset information of the target obstacle if the overlap state between the vehicle and the target obstacle is overlap, the attribute detection result is a partially crushable attribute, and the preset attribute of the target obstacle in the preset information of the target obstacle is a partially crushable attribute, and determining that the perception evaluation result of the vehicle in the attribute dimension is abnormal if the chassis height of the vehicle is less than the height of the target obstacle.

8. The device according to claim 7, wherein The acquisition module is configured to acquire a plurality of detection frames of the vehicle detecting the target obstacle; wherein each of the plurality of detection frames includes a detection result indicating whether the target obstacle is detected; The evaluation module is used to determine at least one of the farthest detection distance, the stable detection distance, the safe detection distance and the frame pass rate based on the multiple detection frames and the preset information of the target obstacle; and determine the perception evaluation result of the vehicle under the missed detection rate dimension based on at least one of the farthest detection distance, the stable detection distance, the safe detection distance and the frame pass rate.

9. The device according to claim 8, wherein The evaluation module is configured to perform at least one of the following: Determining a first valid frame from one or more valid frames included in the multiple detection frames, and determining a first detected position of the vehicle based on a time corresponding to the first valid frame; obtaining the maximum detection distance based on the position of the target obstacle included in the preset information of the target obstacle and the first detected position of the vehicle; wherein each valid frame in the one or more valid frames is a detection frame in which a detection result indicates that the target obstacle is detected; Determining a plurality of consecutive stable detection frames from the plurality of detection frames, and determining a stable detected position of the vehicle based on a time corresponding to a first detection frame among the plurality of consecutive stable detection frames; obtaining the stable detection distance based on the position of the target obstacle included in the preset information of the target obstacle and the stable detected position of the vehicle; wherein the number of consecutive missed detection frames included in the plurality of consecutive stable detection frames is less than a preset threshold value, and the missed detection frame is a detection frame in which a detection result indicates that the target obstacle was not detected; Determining a last detection frame from the multiple detection frames based on a position of the target obstacle included in the preset information of the target obstacle; obtaining a total number of frames between the first valid frame and the last detection frame, and a number of valid frames between the first valid frame and the last detection frame, and obtaining a frame pass rate based on the number of valid frames and the total number of frames; Determine a plurality of consecutive stable detection frames from the plurality of detection frames, determine a reaction distance of the vehicle based on the vehicle speed corresponding to the first stable detection frame in the plurality of consecutive stable detection frames, and obtain the safe detection distance based on the reaction distance of the vehicle and the braking distance of the vehicle.

10. The device according to claim 8, wherein The evaluation module is configured to determine that a perception evaluation result of the vehicle under the missed detection rate dimension is abnormal when a preset condition is determined to be satisfied based on at least one of the maximum detection distance, the stable detection distance, the safe detection distance, and the frame pass rate; Among them, the preset conditions include at least one of the following: the farthest detection distance is less than the preset detection distance, the stable detection distance is less than the preset detection distance, the safe detection distance is less than the preset safety distance, and the frame pass rate is less than the preset frame pass rate.

11. The device according to any one of claims 7 to 10, wherein: The acquisition module is used to obtain the area range detection frame of the vehicle to the target obstacle; The evaluation module is configured to determine a perception evaluation result of the vehicle in the area range dimension based on the area range detection frame, the safety detection frame in the preset information of the target obstacle, and the near-host vehicle detection frame; wherein the safety detection frame includes the area where the target obstacle is located; The near-host-vehicle detection frame includes a portion of the area where the target obstacle is located that is closest to the vehicle.

12. The device according to claim 11, wherein The evaluation module is configured to determine a first intersection and a first union between the area range detection frame and the safety detection frame in the preset information of the target obstacle, and determine a first ratio based on the first intersection and the first union; Determine a second intersection between the area range detection frame and the near-host vehicle detection frame in the preset information of the target obstacle, and determine a second ratio based on the second intersection and the near-host vehicle detection frame; when the first ratio is less than a first threshold and / or the second ratio is less than a second threshold, determine that the perception evaluation result of the vehicle in the area range dimension is abnormal.

13. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 6.

14. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 6.

15. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 6.

Images