Perception data processing method and device, storage medium and electronic equipment

Abstract

The invention provides a sensing data processing method and device, a storage medium and electronic equipment, and the method comprises the steps: obtaining initial sensing data collected by a vehicle, the initial sensing data being edge scene data in the sensing data collected by the vehicle; matching the initial sensing data with a preset driving scene template to obtain driving scene data; generating a target scene file according to the driving scene data; and generalizing the target scene file to generate a plurality of target sensing data, the target sensing data being used for training a sensing algorithm. By converting the initial perception data acquired by the vehicle into the generalizable target scene file and generalizing the initial perception data according to the target scene file, the data volume for training the perception algorithm is enriched, so that the training effect of the perception algorithm is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of intelligent driving technology, and in particular to a method, device, storage medium and electronic device for processing perception data. Background Art

[0002] Perception algorithms are an important part of machine learning and artificial intelligence. In the field of intelligent driving, perception algorithms can be used to understand and interpret the environment around the vehicle. By continuously improving the accuracy and real-time performance of environmental perception, perception algorithms can help autonomous vehicles drive more safely and efficiently.

[0003] In order to ensure the recognition accuracy of the perception algorithm, the perception algorithm needs to be trained. In the field of autonomous driving, the perception data used to train the perception algorithm is generally collected by professional collection vehicles on public roads. However, the number of edge scene data that has a significant effect on improving the perception algorithm is small in the perception data obtained in this way, which easily leads to poor training effects on the perception algorithm. Summary of the invention

[0004] In view of this, the present disclosure provides a method, device, storage medium and electronic device for processing perception data, which can enrich the amount of edge scene data in the perception data and improve the training effect of the perception algorithm.

[0005] According to a first aspect of an embodiment of the present disclosure, a method for processing perception data is provided, the method comprising:

[0006] Acquire initial perception data collected by the vehicle, wherein the initial perception data is edge scene data in the perception data collected by the vehicle;

[0007] Matching the initial perception data with a preset driving scene template to obtain driving scene data;

[0008] generating a target scene file according to the driving scene data;

[0009] The target scene file is generalized to generate a plurality of target perception data, wherein the target perception data is used to train a perception algorithm.

[0010] In one embodiment, matching the initial perception data with a preset driving scene template to obtain driving scene data includes:

[0011] Restoring the trajectories of multiple dynamic elements in the initial perception data, wherein the dynamic elements are target objects in motion in the initial perception data;

[0012] Splitting the trajectories of the multiple dynamic elements to obtain the trajectory of each dynamic element;

[0013] For each dynamic element, matching the trajectory of the dynamic element with a plurality of driving scene templates respectively, and determining a target driving scene template that successfully matches the trajectory of the dynamic element from the plurality of driving scene templates;

[0014] The driving scene data is determined according to the target driving scene templates of the multiple dynamic elements.

[0015] In one embodiment, matching the trajectory of the dynamic element with a plurality of driving scene templates respectively, and determining a target driving scene template that successfully matches the trajectory of the dynamic element from the plurality of driving scene templates, comprises:

[0016] Based on the driving scenario template, determining the constraint conditions corresponding to each driving scenario;

[0017] The trajectory of each dynamic element is matched with the constraint conditions corresponding to different driving scenes respectively, and the target driving scene template corresponding to each dynamic element is determined.

[0018] In one embodiment, the initial sensing data carries a timestamp;

[0019] The step of generating a target scene file according to the driving scene data comprises:

[0020] Determining the time sequence of the target driving scene template of each dynamic element in the driving scene data according to the timestamp in the initial perception data;

[0021] Determining an arrangement of multiple target driving scenes in the driving scene data based on a time sequence of the target driving scene template of each dynamic element;

[0022] generating an initial scenario file based on the arrangement of the plurality of target driving scenarios;

[0023] Insert high-precision map data into the initial scene file to generate the target scene file.

[0024] In one embodiment, generalizing the target scene file to generate a plurality of target perception data includes:

[0025] Based on a preset scene mapping relationship, generalizing the dynamic elements in the target scene file by using a generalization tool, wherein the scene mapping relationship is a mapping relationship between the dynamic elements and the driving scene;

[0026] A plurality of target perception data are generated according to the generalization result of the dynamic element.

[0027] In one embodiment, the method further comprises:

[0028] According to the types of preset driving scenes, determining the constraint conditions of the dynamic elements in each type of driving scene;

[0029] Construct each type of driving scenario template based on the constraints of dynamic elements.

[0030] According to a second aspect of an embodiment of the present disclosure, a device for processing perception data is provided, the device comprising:

[0031] An acquisition unit, configured to acquire initial perception data collected by the vehicle, wherein the initial perception data is edge scene data in the perception data collected by the vehicle;

[0032] A matching unit, used for matching the initial perception data with a preset driving scene template to obtain driving scene data;

[0033] A generating unit, used for generating a target scene file according to the driving scene data;

[0034] A generalization unit is used to generalize the target scene file to generate a plurality of target perception data, wherein the target perception data is used to train the perception algorithm.

[0035] In one embodiment, the matching unit is used to:

[0036] Restoring the trajectories of multiple dynamic elements in the initial perception data, wherein the dynamic elements are target objects in motion in the initial perception data;

[0037] Splitting the trajectories of the multiple dynamic elements to obtain the trajectory of each dynamic element;

[0038] For each dynamic element, matching the trajectory of the dynamic element with a plurality of driving scene templates respectively, and determining a target driving scene template that successfully matches the trajectory of the dynamic element from the plurality of driving scene templates;

[0039] The driving scene data is determined according to the target driving scene templates of the multiple dynamic elements.

[0040] In one embodiment, the matching unit is used to:

[0041] Based on the driving scenario template, determining the constraint conditions corresponding to each driving scenario;

[0042] The trajectory of each dynamic element is matched with the constraint conditions corresponding to different driving scenes respectively, and the target driving scene template corresponding to each dynamic element is determined.

[0043] In one embodiment, the initial sensing data carries a timestamp;

[0044] The generating unit is used for:

[0045] Determining the time sequence of the target driving scene template of each dynamic element in the driving scene data according to the timestamp in the initial perception data;

[0046] Determining an arrangement of multiple target driving scenes in the driving scene data based on a time sequence of the target driving scene template of each dynamic element;

[0047] generating an initial scenario file based on the arrangement of the plurality of target driving scenarios;

[0048] Insert high-precision map data into the initial scene file to generate the target scene file.

[0049] In one embodiment, the generalization unit is used to:

[0050] Based on a preset scene mapping relationship, generalizing the dynamic elements in the target scene file by using a generalization tool, wherein the scene mapping relationship is a mapping relationship between the dynamic elements and the driving scene;

[0051] A plurality of target perception data are generated according to the generalization result of the dynamic element.

[0052] In one embodiment, the apparatus further comprises:

[0053] A determination unit, used to determine the constraint conditions of the dynamic elements in each type of driving scene according to the type of preset driving scene;

[0054] The construction unit is used to construct each type of driving scene template according to the constraints of the dynamic elements.

[0055] According to a third aspect of an embodiment of the present disclosure, there is provided a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the steps of any method described in the first aspect are implemented.

[0056] According to a fourth aspect of an embodiment of the present disclosure, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of any method described in the first aspect are implemented.

[0057] According to a fifth aspect of an embodiment of the present disclosure, a computer program product is provided, which, when executed by a processor, implements the steps of any method described in the first aspect.

[0058] The technical solution provided by the embodiments of the present disclosure may have the following beneficial effects:

[0059] By converting the initial perception data collected by the vehicle into a generalizable target scene file and generalizing the initial perception data according to the target scene file, the amount of data that can be used to train the perception algorithm is enriched, thereby improving the training effect of the perception algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 is a flow chart of a method for processing perception data shown in an exemplary embodiment of the present disclosure;

[0061] Figure 2 is a flowchart of constructing a driving scene template shown in an exemplary embodiment of the present disclosure;

[0062] Figure 3 is a flow chart of determining driving scene data shown in an exemplary embodiment of the present disclosure;

[0063] Figure 4 is a flow chart showing a method of generating a target scene file according to an exemplary embodiment of the present disclosure;

[0064] Figure 5 An exemplary embodiment of the present disclosure shows a flowchart for generalizing a target scene file;

[0065] Figure 6 is a block diagram of a device for processing perception data shown in an exemplary embodiment of the present disclosure;

[0066] Figure 7 It is a hardware structure diagram of an electronic device shown in an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION

[0067] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0068] The terms used in this disclosure are for the purpose of describing specific embodiments only and are not intended to limit the disclosure. The singular forms of "a", "said" and "the" used in this disclosure and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0069] It should be understood that although the terms first, second, third, etc. may be used in the present disclosure to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present disclosure, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0070] Perception algorithms are an important part of machine learning and artificial intelligence. In the field of intelligent driving, perception algorithms can be used to understand and interpret the environment around the vehicle. By continuously improving the accuracy and real-time performance of environmental perception, perception algorithms can help autonomous vehicles drive more safely and efficiently.

[0071] In order to improve the perception algorithm's accuracy in perceiving the vehicle's surroundings, it is necessary to use perception data to train the perception algorithm. However, since the number of general scenes accounts for the majority of the perception data collected by vehicles on public roads, it is not very valuable for the training of the perception algorithm. The number of edge scene data that is very helpful for improving the perception algorithm is small, which leads to the problem of poor training effect of the perception algorithm.

[0072] Based on this, the present disclosure provides a method for processing perception data, which can be applied to electronic devices such as computers. For the specific process, please refer to Figure 1 The flowchart shown includes the following steps:

[0073] S101, obtaining initial perception data collected by the vehicle.

[0074] The initial perception data is the edge scene data in the perception data collected by the vehicle. The edge scene data refers to the perception data of driving scenes with a low probability of occurrence collected while the vehicle is driving. For example, the perception data obtained under extreme weather conditions such as heavy rain, blizzard, haze, etc., the perception data in these situations may be completely different from the situation in normal weather. For another example, at night or in an environment with extremely weak light, the data collected by the physical sensors in the vehicle may be very different from the normal light intensity data. Alternatively, in extreme situations such as traffic accident scenes and traffic light failures, the perception data collected by the vehicle will also be different from normal situations.

[0075] S102, matching the initial perception data with a preset driving scene template to obtain driving scene data.

[0076] The preset driving scene template is a template designed in advance according to the vehicle driving scene, and is used to describe different vehicle driving scenes. For example, with respect to the vehicle driving scene, the template can be designed according to driving scenes such as vehicle cutting in, vehicle cutting out, and vehicle driving in the opposite direction. For another example, with respect to the driving scene of pedestrians, the template can be designed according to driving scenes such as pedestrians crossing. Or, with respect to the driving scene of bicycles, the template can be designed according to driving scenes such as bicycles crossing.

[0077] Please refer to the construction process of the driving scene template for details. Figure 2 The flowchart shown includes the following steps:

[0078] S201, determining the constraint conditions of the dynamic elements in each type of driving scene according to the type of preset driving scenes.

[0079] S202: construct a driving scene template of each type according to the constraints of the dynamic elements.

[0080] When constructing a driving scene template, it is necessary to first determine the constraints of the dynamic elements in each driving scene according to the type of preset driving scenes, such as vehicle cut-in, vehicle cut-out, vehicle driving in opposite directions, etc. The constraints include at least a start time constraint, an end time constraint, and a process constraint. Among them, the start time constraint is used to determine the state of the dynamic element at the start time of the driving scene, the end time constraint is used to determine the state of the dynamic element at the end time of the driving scene, and the process constraint is used to determine the state of the dynamic element between the start time and the end time in the driving scene. The constraints may include: heading angle information, coordinate information of the dynamic element compared to the main vehicle, and lane information.

[0081] Then, according to the constraints, each type of driving scenario template can be constructed. By combining the driving scenario templates, an OSC2.0 (OpenScenario2.0, an open standard) file can be generated, and then complex and realistic traffic scenarios can be generated. These scenarios can help the perception algorithm to be trained under various conditions, thereby improving the robustness and accuracy of the algorithm, thereby ensuring the reliability of the perception algorithm in the real world and promoting the development of autonomous driving technology.

[0082] After the initial perception data is acquired, the perception data can be matched according to the driving scene described by the driving scene template. In one embodiment, the driving trajectory of other vehicles in the perception data can be matched with the driving scene template about the vehicle in the driving scene template to determine the driving scene template corresponding to the other vehicles. The trajectory includes at least: the trajectory point, timestamp information, orientation angle information, coordinate information of the dynamic element compared to the main vehicle, and lane information. For example, if the driving trajectory of other vehicles successfully matches the conditions corresponding to the vehicle cutting into the driving template, then the vehicle corresponds to the vehicle cutting into the driving scene template.

[0083] After determining the driving scene templates corresponding to the multiple dynamic elements in the initial perception data, the multiple driving scene templates are the driving scene data.

[0084] In one embodiment, the process of determining driving scene data can be found in Figure 3 The flowchart shown includes the following steps:

[0085] S301, restoring trajectories of multiple dynamic elements in the initial perception data, wherein the dynamic elements are target objects in motion in the initial perception data.

[0086] S302: Split the trajectories of the multiple dynamic elements to obtain the trajectory of each dynamic element.

[0087] S303: For each dynamic element, respectively match the trajectory of the dynamic element with a plurality of driving scene templates, and determine a target driving scene template that successfully matches the trajectory of the dynamic element from the plurality of driving scene templates.

[0088] S304: Determine the driving scene data according to the target driving scene templates of the multiple dynamic elements.

[0089] After the initial perception data is obtained, the position and posture information of the main vehicle can be calculated based on the initial perception data. Then the dynamic elements (for example, vehicles, pedestrians, bicycles, etc.) in the initial perception data are identified, and the position and posture information and coordinate information of the dynamic elements relative to the main vehicle are calculated. According to the position and posture information of the main vehicle and the position and coordinate information of the dynamic elements, the trajectory of the dynamic elements can be restored. For example, the image sequence data in the initial perception data collected by the vehicle is used to calculate the precise position and posture information of the main vehicle using the SLAM (Simultaneous Localization and Mapping) algorithm. Alternatively, GPS (Global Positioning System) data and IMU (Inertial Measurement Unit) data are collected by the vehicle, and the position and posture information of the main vehicle is calculated based on the GPS data and IMU data.

[0090] For the initial perception data collected by the same vehicle, there may be multiple trajectories of dynamic elements. When restoring the dynamic element trajectory based on the initial perception data, the trajectories of multiple dynamic elements are restored together. Therefore, if you want to determine the driving scene of each dynamic element trajectory, you need to split the trajectory of each dynamic element and then analyze each dynamic element separately. For example, the dynamic elements in the initial perception data are vehicles and pedestrians. After analyzing the image sequence data, the motion trajectories of all dynamic elements are obtained. Then, in the dynamic element trajectory set, the vehicle trajectory and pedestrian trajectory are split out to determine the driving scene template corresponding to the vehicle and the driving scene template corresponding to the pedestrian.

[0091] After obtaining the trajectory of each dynamic element, for each dynamic element, the trajectory of the dynamic element is matched with multiple driving scene templates respectively. Specifically, the corresponding driving scene template of the same category can be determined according to the category of the dynamic element to narrow the matching range of the driving scene template and improve the efficiency of determining the target driving scene template. For example, when the dynamic element is a vehicle, its corresponding driving scene template of the same category may include a vehicle entry scene template, a vehicle exit scene template, or a vehicle oncoming driving scene template. After determining the driving scene template of the same category, the trajectory of the dynamic element in each driving scene template is converted into a constraint condition. Through the constraint condition, it is determined whether the trajectory of the dynamic element in the initial perception data is consistent with the dynamic trajectory in the driving scene template. If it is consistent, the driving scene template is determined as the target driving scene template.

[0092] Furthermore, for the same dynamic element, if its changes are more complex, there may be a situation where one dynamic element matches multiple driving scene templates. In this case, the matching rate can be calculated based on the matching degree between the trajectory points in the dynamic element trajectory and the process constraints of each driving scene template. Then, a preferred option is determined from the multiple matching driving scene templates as the target driving scene template based on the matching rate.

[0093] Finally, according to the target scene template corresponding to the multiple dynamic elements in the initial perception data, the driving scene data corresponding to the initial perception data can be determined.

[0094] S103: Generate a target scene file according to the driving scene data.

[0095] After obtaining the driving scene data, multiple target scene templates in the driving scene data are sorted to generate a target scene file. In one embodiment, the arrangement order of the target scene templates corresponding to each dynamic element in the target scene file can be determined according to the time information in the initial perception data, and then the multiple target scene templates can be arranged according to the arrangement order to generate a target scene file, for example, an OSC2.0 file.

[0096] In some embodiments, the process of generating the target scene file is as follows: Figure 4 As shown, Figure 4 FIG. 1 is a flowchart of generating a target scene file exemplarily shown in an embodiment of the present disclosure, and the flowchart includes the following steps:

[0097] S401, determining the time sequence of the target driving scene template of each dynamic element in the driving scene data according to the timestamp in the initial perception data.

[0098] S402: Determine an arrangement of multiple target driving scenes in the driving scene data based on the time sequence of the target driving scene template of each dynamic element.

[0099] S403: Generate an initial scenario file based on the arrangement of the multiple target driving scenarios.

[0100] S404: Insert high-precision map data into the initial scene file to generate the target scene file.

[0101] When the physical sensor of the vehicle collects the initial perception data, the initial perception data already carries timestamp information, which is used to indicate the time when the perception data was collected. The order of arranging the target driving scenes corresponding to the multiple dynamic elements in the initial perception data is based on the timestamp in the initial perception data. Based on the timestamp in the initial perception data, the start time corresponding to the target driving scene template of each dynamic element in the driving scene data can be determined, and the time order of the multiple target driving scenes can be determined based on the order of the start times.

[0102] After determining the time sequence, the arrangement of the target driving scene templates corresponding to multiple dynamic elements in the driving scene data can be determined according to the time sequence of the target driving scene templates of each dynamic element. For example, if the start time of the target driving scene templates corresponding to three dynamic elements is the same, that is, the time sequence of the three target driving scene templates is parallel, then it can be determined that the arrangement of the target driving scenes corresponding to the multiple dynamic elements is a parallel arrangement. If there is a sequence in the start time of the target driving scene templates corresponding to two dynamic elements in the initial perception data, that is, the time sequence of the two target driving scene templates is serial, then it can be determined that the arrangement of the target driving scenes corresponding to the two dynamic elements is a serial arrangement.

[0103] After the target driving scene is arranged according to the arrangement mode of the target driving scene template, the initial scene file corresponding to the initial perception data can be generated, and the initial scene file can be used to describe the trajectory changes of other dynamic elements compared with the main vehicle during the driving process of the main vehicle. After the initial scene file is generated, in order to further improve the information in the initial scene file, the high-precision map information can be inserted into the initial scene file. The high-precision map is used to provide detailed road information, traffic signs, lane structure and other data, which can increase the realism and accuracy of the simulation scene in the scene file. In addition, weather information, environmental information and other information can also be added to the initial scene file to simulate factors that affect vehicle driving safety, such as extreme weather (such as heavy rain, blizzard, haze, etc.) and different light environments. After the target scene file is generated, the target scene file can be used to train the perception algorithm to improve the accuracy and adaptability of the algorithm and ensure the safety and reliability of the autonomous driving system in the real world environment.

[0104] S104, generalizing the target scene file to generate a plurality of target perception data.

[0105] After obtaining the target scene file, the target scene file can be imported into the simulation platform, and the generalization tool is called through the simulation platform to generalize the target scene file, thereby generating perception-level cases for training the perception algorithm, including images, annotated true value data and other target perception data.

[0106] In one embodiment, the process of generalizing the target scene file is shown in Figure 5 The flowchart shown includes the following steps:

[0107] S501 , generalizing the dynamic elements in the target scene file by using a generalization tool based on a preset scene mapping relationship.

[0108] S502: Generate a plurality of target perception data according to the generalization result of the dynamic element.

[0109] The scene mapping relationship is a mapping relationship between dynamic elements and driving scenes. According to the scene mapping relationship, the generalization tool can generalize the dynamic elements in the target scene file. For example, if the category of the dynamic elements in the initial perception data is vehicle, then according to the scene mapping relationship, the color, brand and other features of the vehicle can be generalized to obtain the driving scene data of vehicles of different colors and brands after generalization. By generalizing the driving scene, the data information of the edge data can be further enriched. This is because under dim light conditions, some colors (such as white or bright vehicles) are easier to identify than dark vehicles, which may result in different reaction times for the main vehicle. For another example, different vehicle brands have different vehicle parameters. By generalizing the vehicle brand, the driving scenes of vehicles of different brands on real roads can be simulated.

[0110] After generalizing the dynamic elements, various types of data can be generated for training perception algorithms, including images, labeled true value data, and other target perception data.

[0111] By converting the initial perception data collected by the vehicle into a generalizable target scene file and generalizing the initial perception data according to the target scene file, the amount of data that can be used to train the perception algorithm is enriched, thereby improving the training effect of the perception algorithm.

[0112] Corresponding to the embodiments of the aforementioned method for processing perception data, the present disclosure also provides embodiments of an apparatus for processing perception data.

[0113] Please refer to Figure 6 , Figure 6 is a block diagram of a device for processing perception data shown in an exemplary embodiment of the present disclosure, the device comprising:

[0114] An acquisition unit 601 is used to acquire initial perception data collected by the vehicle, wherein the initial perception data is edge scene data in the perception data collected by the vehicle;

[0115] A matching unit 602 is used to match the initial perception data with a preset driving scene template to obtain driving scene data;

[0116] A generating unit 603, configured to generate a target scene file according to the driving scene data;

[0117] The generalization unit 604 is used to generalize the target scene file to generate a plurality of target perception data, wherein the target perception data is used to train the perception algorithm.

[0118] In one embodiment, the matching unit 602 is used to:

[0119] Restoring the trajectories of multiple dynamic elements in the initial perception data, wherein the dynamic elements are target objects in motion in the initial perception data;

[0120] Splitting the trajectories of the multiple dynamic elements to obtain the trajectory of each dynamic element;

[0121] For each dynamic element, matching the trajectory of the dynamic element with a plurality of driving scene templates respectively, and determining a target driving scene template that successfully matches the trajectory of the dynamic element from the plurality of driving scene templates;

[0122] The driving scene data is determined according to the target driving scene templates of the multiple dynamic elements.

[0123] In one embodiment, the matching unit 602 is used to:

[0124] Based on the driving scenario template, determining the constraint conditions corresponding to each driving scenario;

[0125] The trajectory of each dynamic element is matched with the constraint conditions corresponding to different driving scenes respectively, and the target driving scene template corresponding to each dynamic element is determined.

[0126] In one embodiment, the initial sensing data carries a timestamp;

[0127] The generating unit 603 is used for:

[0128] Determining the time sequence of the target driving scene template of each dynamic element in the driving scene data according to the timestamp in the initial perception data;

[0129] Determining an arrangement of multiple target driving scenes in the driving scene data based on a time sequence of the target driving scene template of each dynamic element;

[0130] generating an initial scenario file based on the arrangement of the plurality of target driving scenarios;

[0131] Insert high-precision map data into the initial scene file to generate the target scene file.

[0132] In one embodiment, the generalization unit 604 is used to:

[0133] Based on a preset scene mapping relationship, generalizing the dynamic elements in the target scene file by using a generalization tool, wherein the scene mapping relationship is a mapping relationship between the dynamic elements and the driving scene;

[0134] A plurality of target perception data are generated according to the generalization result of the dynamic element.

[0135] In one embodiment, the apparatus further comprises:

[0136] A determination unit, used to determine the constraint conditions of the dynamic elements in each type of driving scene according to the type of preset driving scene;

[0137] The construction unit is used to construct each type of driving scene template according to the constraints of the dynamic elements.

[0138] The implementation process of the functions and effects of each unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0139] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can refer to the partial description of the method embodiments. The device embodiments described above are only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0140] The embodiments of the processing device for perceptual data disclosed in the present invention can be applied to electronic devices such as computers. The device embodiments can be implemented by software, hardware, or a combination of software and hardware. Taking software implementation as an example, as a device in a logical sense, it is formed by the processor of the electronic device in which it is located reading the corresponding computer program instructions in the non-volatile memory into the memory and running them. From the hardware level, if Figure 7 As shown, it is a hardware structure diagram of an electronic device where the processing device of the perception data disclosed in the present invention is located, except Figure 7 In addition to the processor, memory, network interface, and non-volatile memory shown, the electronic device in which the device in the embodiment is located may also include other hardware according to its actual functions, which will not be described in detail.

[0141] Based on the same inventive concept, the present disclosure also provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the steps of the above-mentioned method for processing perception data are implemented.

[0142] Based on the same inventive concept, the present disclosure also provides a computer program product on which a computer program is stored. When the computer program product is executed by a processor, the steps of the above-mentioned method for processing perception data are implemented.

[0143] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims can be performed in a different order and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the particular order or sequential order shown to achieve the desired results. In some implementations, multitasking and parallel processing may be advantageous.

[0144] The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present disclosure shall be included in the scope of protection of the present disclosure.

Claims

1. A method for processing perception data, characterized in that: The method comprises: Acquire initial perception data collected by the vehicle, wherein the initial perception data is edge scene data in the perception data collected by the vehicle; Matching the initial perception data with a preset driving scene template to obtain driving scene data; generating a target scene file according to the driving scene data; The target scene file is generalized to generate a plurality of target perception data, wherein the target perception data is used to train a perception algorithm.

2. The method according to claim 1, characterized in that The matching of the initial perception data with a preset driving scene template to obtain driving scene data includes: Restoring the trajectories of multiple dynamic elements in the initial perception data, wherein the dynamic elements are target objects in motion in the initial perception data; Splitting the trajectories of the multiple dynamic elements to obtain the trajectory of each dynamic element; For each dynamic element, matching the trajectory of the dynamic element with a plurality of driving scene templates respectively, and determining a target driving scene template that successfully matches the trajectory of the dynamic element from the plurality of driving scene templates; The driving scene data is determined according to the target driving scene templates of the multiple dynamic elements.

3. The method according to claim 2, characterized in that The step of matching the trajectory of the dynamic element with a plurality of driving scene templates respectively and determining a target driving scene template that successfully matches the trajectory of the dynamic element from the plurality of driving scene templates comprises: Based on the driving scenario template, determining the constraint conditions corresponding to each driving scenario; The trajectory of each dynamic element is matched with the constraint conditions corresponding to different driving scenes respectively, and the target driving scene template corresponding to each dynamic element is determined.

4. The method according to claim 1, characterized in that: The initial sensing data carries a timestamp; The step of generating a target scene file according to the driving scene data comprises: Determining the time sequence of the target driving scene template of each dynamic element in the driving scene data according to the timestamp in the initial perception data; Determining an arrangement of multiple target driving scenes in the driving scene data based on a time sequence of the target driving scene template of each dynamic element; generating an initial scenario file based on the arrangement of the plurality of target driving scenarios; Insert high-precision map data into the initial scene file to generate the target scene file.

5. The method according to claim 1, characterized in that The generalizing the target scene file to generate a plurality of target perception data includes: Based on a preset scene mapping relationship, generalizing the dynamic elements in the target scene file by using a generalization tool, wherein the scene mapping relationship is a mapping relationship between the dynamic elements and the driving scene; A plurality of target perception data are generated according to the generalization result of the dynamic element.

6. The method according to claim 1, characterized in that The method further comprises: According to the types of preset driving scenes, determining the constraint conditions of the dynamic elements in each type of driving scene; Construct each type of driving scenario template based on the constraints of dynamic elements.

7. A sensory data processing device, characterized in that: The device comprises: An acquisition unit, configured to acquire initial perception data collected by the vehicle, wherein the initial perception data is edge scene data in the perception data collected by the vehicle; A matching unit, used for matching the initial perception data with a preset driving scene template to obtain driving scene data; A generating unit, used for generating a target scene file according to the driving scene data; A generalization unit is used to generalize the target scene file to generate a plurality of target perception data, wherein the target perception data is used to train the perception algorithm.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that When the computer program product is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images