Driving task processing method, device, electronic device and storage medium

Abstract

The present invention relates to the field of image recognition technology and provides a driving task processing method, apparatus, electronic device, and storage medium. The method comprises: obtaining image features of an image to be tested; inputting the image features of the image to be tested into a processing model to obtain task attributes of a task target in the image to be tested, output by the processing model; the task attributes represent the perception results of the task target under different perception tasks; and the processing model is a model obtained through machine learning training, using image features in an image sample as input and the perception results of the task target in the image sample under various perception tasks as output, for determining the task attributes in the image. The driving task processing method, system, electronic device, and storage medium provided by the present invention implement branching processing of the image to be tested, enabling each perception task to share a single processing model, thereby reducing the need to establish multiple models for different perception tasks, lowering overall system memory usage, and reducing workload.

Description

Technical Field

[0001] The present invention relates to the field of intelligent vehicle driving technology, and in particular to a driving task processing method, system, electronic device and storage medium. Background Art

[0002] The perception tasks in autonomous driving scenarios are complex and diverse, including 2D / 3D object detection, general obstacle detection, lane detection, traffic sign detection, road marking detection, key point detection, drivable area segmentation, monocular depth estimation, optical flow estimation, and re-identification. Existing methods typically train a separate network for each perception task or category, and then deploy each one separately on the product side. This results in a large number of models, high memory usage, high computational overhead, and heavy deployment effort. Summary of the Invention

[0003] In response to the problems existing in the prior art, the present invention provides a driving task processing method, system, electronic device and storage medium.

[0004] In a first aspect, the present invention provides a driving task processing method, comprising:

[0005] Obtaining image features of the image to be tested;

[0006] Inputting the image features of the image to be tested into a processing model to obtain task attributes of the task target in the image to be tested output by the processing model; the task attributes represent the perception results of the task target under different perception tasks;

[0007] The processing model is a model obtained through machine learning training that uses image features in image samples as input and the perception results of task targets in image samples under various perception tasks as output, and is used to determine task attributes in images.

[0008] The processing model includes a detection layer and a processing layer. The detection layer is used to determine the judgment parameters of the task target in the image to be tested according to the image features of the image to be tested; the processing layer is used to determine the task attributes of the task target according to the assigned judgment parameters.

[0009] In one embodiment, the image features of the image to be tested are input into a processing model to obtain the task attributes of the task target in the image to be tested output by the processing model, including:

[0010] Inputting the image features of the image to be tested into the detection layer, the detection layer determines the detection rules corresponding to each perception task type according to a plurality of pre-stored perception task types, and determines the judgment parameters of the task targets corresponding to each perception task type according to the determined plurality of detection rules and the image features of the image to be tested;

[0011] The judgment parameters of the task targets corresponding to each perception task type are input into the processing layer, and the processing layer determines the judgment rules corresponding to each perception task type based on multiple pre-stored perception task types. The judgment is performed based on the determined multiple judgment rules and the judgment parameters corresponding to the multiple judgment rules to obtain the task attributes of the task targets.

[0012] In one embodiment, the method further comprises: obtaining correlations between the perception task types;

[0013] Accordingly, the image features of the image to be tested are input into the processing model to obtain the task attributes of the task target in the image to be tested output by the processing model, including:

[0014] Inputting the image features of the image to be tested into the detection layer, the detection layer determines the detection rules corresponding to each perception task type according to a plurality of pre-stored perception task types, and determines the judgment parameters of the task targets corresponding to each perception task type according to the determined plurality of detection rules and the image features of the image to be tested;

[0015] The judgment parameters of the task targets corresponding to each perception task type are input into the processing layer, and the processing layer determines the first type of perception task type and the second type of perception task type according to the correlation between the perception task types, determines the corresponding judgment rules according to the first type of perception task type and the second type of perception task type, and divides the judgment parameters of the task targets corresponding to each perception task type to obtain new judgment parameters, and makes judgments based on the determined multiple judgment rules and the new judgment parameters corresponding to the multiple judgment rules to obtain the task attributes of the task targets.

[0016] In one embodiment, the distribution layer includes a partitioning layer and a plurality of sub-processing layers;

[0017] Accordingly, the division layer determines the first and second types of perception task types according to the correlation between the perception task types, determines the corresponding judgment rules according to the first and second types of perception task types, and divides the judgment parameters of the task targets corresponding to each perception task type to obtain new judgment parameters;

[0018] The sub-processing layer performs judgment according to the assigned judgment rule and the corresponding new judgment parameter to obtain the task attributes of the task target.

[0019] In one embodiment, the perception tasks include 2D / 3D target detection, obstacle detection, lane line detection, traffic sign detection, road marking detection, key point detection, drivable area segmentation, monocular depth estimation, and optical flow estimation.

[0020] In a second aspect, the present invention provides a driving task processing device, comprising:

[0021] An acquisition module, used to obtain image features of the image to be tested;

[0022] a processing module, configured to input the image features of the image to be tested into a processing model to obtain task attributes of the task target in the image to be tested output by the processing model; the task attributes represent the perception results of the task target under different perception tasks;

[0023] The processing model is a model obtained through machine learning training that uses image features in image samples as input and the perception results of task targets in image samples under various perception tasks as output, and is used to determine task attributes in images.

[0024] The processing model includes a detection layer and a processing layer. The detection layer is used to determine the judgment parameters of the task target in the image to be tested according to the image features of the image to be tested; the processing layer is used to determine the task attributes of the task target according to the assigned judgment parameters.

[0025] In a third aspect, the present invention provides an electronic device comprising a memory and a memory storing a computer program, wherein the processor implements the steps of the driving task processing method described in the first aspect when executing the program.

[0026] In a fourth aspect, the present invention provides a processor-readable storage medium, wherein the processor-readable storage medium stores a computer program, and the computer program is used to enable the processor to execute the steps of the driving task processing method described in the first aspect.

[0027] The driving task processing method, system, electronic device, and storage medium provided by the present invention implement branch processing of the images to be tested, thereby enabling each perception task to share a processing model, reducing the need to establish multiple models for different perception tasks, lowering the overall system memory usage, and reducing the workload, thereby achieving the goals of strong scalability and easy maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0029] Figure 1 is a flow chart of the driving task processing method provided by the present invention;

[0030] Figure 2It is a structural schematic diagram of the driving task processing device provided by the present invention;

[0031] Figure 3 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0032] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0033] The following combination Figure 1-Figure 3 The present invention describes a driving task processing method, device, electronic device, and storage medium.

[0034] Figure 1 A schematic diagram of a driving task processing method of the present invention is shown. Figure 1 , the method comprising:

[0035] 11. Obtain image features of the image to be tested;

[0036] 12. Input the image features of the image to be tested into the processing model to obtain the task attributes of the task target in the image to be tested output by the processing model; the task attributes represent the perception results of the task target under different perception tasks; wherein the processing model is a model trained through machine learning, which uses the image features in the image sample as input and the perception results of the task target in the image sample under various perception tasks as output, and is used to determine the task attributes in the image;

[0037] The processing model includes a detection layer and a processing layer. The detection layer is used to determine the judgment parameters of the task target in the image to be tested according to the image features of the image to be tested; the processing layer is used to determine the task attributes of the task target according to the assigned judgment parameters.

[0038] Regarding steps 11 and 12, an intelligent vehicle with autonomous driving capabilities must perceive the driving conditions in front and around it during driving to obtain perception results, so that it can perform driving maneuvers based on the perception results and enhance safety. To this end, multiple perception tasks need to be added during the vehicle's autonomous driving process to obtain more comprehensive perception results. Perception tasks may include, but are not limited to, 2D / 3D target detection, obstacle detection, lane detection, traffic sign detection, road marking detection, key point detection, drivable area segmentation, monocular depth estimation, and optical flow estimation.

[0039] In the present invention, for the added perception tasks, images in front of the intelligent vehicle can be acquired, and the images can be analyzed and processed to obtain perception results under each perception task.

[0040] First, the collected image is analyzed to obtain image features. The pixels of the image can be used as features.

[0041] In the present invention, the image features of the image to be tested are input into the processing model to obtain the task attributes of the task target in the image to be tested output by the processing model; the task attributes represent the perception results of the task target under different perception tasks.

[0042] To obtain the perception results under each perception task, it is necessary to analyze and process the same image to be tested. Therefore, the task objectives under different perception tasks may be different.

[0043] For example, the perception task is obstacle detection, and its task targets are obstacles, such as other vehicles in front of the vehicle, pedestrians, bicycles, fallen trees, etc.

[0044] For example, the perception task is lane line detection, and its task target is the lane line, such as a single line or double line on the road.

[0045] For example, the perception task is traffic sign detection, and its task target is a traffic sign, such as a traffic sign erected in front of the vehicle.

[0046] It can be seen from this that in the present invention, the image features of the image to be tested are input into the processing model, and the processing model will perform multi-channel analysis on the image features of the image to be tested according to the types of different perception tasks. Each channel corresponds to a perception task, and the detection of each perception task can be performed in parallel to obtain the perception results corresponding to each perception task.

[0047] In the present invention, the processing model is a model obtained through machine learning training, which takes the image features in the image sample as input and the perception results of the task target in the image sample under each perception task as output, and is used to judge the task attributes in the image.

[0048] It should be noted that the driving conditions ahead of different vehicles can be selected and captured using a camera to obtain a large number of image samples. The task targets for each perception task are annotated in these image samples, and the image features of each task target are obtained. The image features are used as input, and the perception results corresponding to the task targets are used as output. This is then trained in an initial neural network model to obtain a processing model.

[0049] In the present invention, the processing model can be updated and upgraded.

[0050] A further method of the present invention is to simultaneously identify the task objectives of each perception task based on image features through the trained model, which is conducive to quickly marking the possibility of task attributes in the image.

[0051] The processing model includes a detection layer and a processing layer. The detection layer is used to determine the judgment parameters of the task target in the image to be tested according to the image features of the image to be tested; the processing layer is used to determine the task attributes of the task target according to the assigned judgment parameters.

[0052] It should be noted that, in the present invention, to obtain the perception results for each perception task, the same image to be tested must be analyzed and processed. Therefore, the task objectives for different perception tasks may differ, as may the judgment parameters for these tasks. Therefore, the image features of the image to be tested are input into the detection layer, which then analyzes the image features of the image to be tested under the constraints of different perception tasks to obtain the corresponding judgment parameters.

[0053] For example, the perception task is whether the vehicle changes lanes, and the parameters generated can be the landing point of the vehicle and the position of the lane line.

[0054] After the judgment parameters of the task targets under each perception task are obtained, the judgment parameters are input into the processing layer, and the processing layer judges the judgment parameters corresponding to each perception task to determine the task attributes of the task targets.

[0055] For example, the positional relationship and / or distance between the landing point of the vehicle and the lane line can be used to determine whether the vehicle has a lane change tendency, is in the process of changing lanes, or has crossed the lane line. Here, whether the vehicle has a lane change tendency, is in the process of changing lanes, or has crossed the lane line is a task attribute of the vehicle.

[0056] The driving task processing method provided by the present invention implements branch processing of the images to be tested, so that each perception task shares a processing model, reduces the need to establish multiple models for different perception tasks, reduces the overall system memory usage, reduces the workload, and achieves the purpose of strong scalability and easy maintenance.

[0057] In the further description of the above method, the processing process of inputting the image features of the image to be tested into the processing model and obtaining the task attributes of the task target in the image to be tested output by the processing model is mainly explained as follows:

[0058] The image features of the image to be tested are input into the detection layer, which determines the detection rules corresponding to each perception task type based on multiple pre-stored perception task types, and determines the judgment parameters of the task targets corresponding to each perception task type based on the determined multiple detection rules and the image features of the image to be tested;

[0059] The judgment parameters of the task targets corresponding to each perception task type are input into the processing layer, which determines the judgment rules corresponding to each perception task type based on multiple pre-stored perception task types, and makes judgments based on the determined multiple judgment rules and the judgment parameters corresponding to the multiple judgment rules to obtain the task attributes of the task targets.

[0060] In this regard, it should be noted that in the present invention, in order to obtain the perception results under each perception task, it is necessary to analyze and process the same image to be tested. For this reason, the task objectives under different perception tasks may be different, and there may also be differences in the judgment parameters of the task objectives. In addition, the rules for analyzing and detecting image features used in the process of obtaining the judgment parameters are also different. Therefore, the image features of the image to be tested are input into the detection layer. Under the constraints of different perception tasks, there are types of perception tasks. Different perception task types correspond to different detection rules, which limit the analysis and calculation process of image features.

[0061] For example, if the perception task is to determine whether a vehicle is changing lanes, the corresponding detection rules include lane line recognition and vehicle landing point determination, such as pixel features of the white area on the lane and the lowest point where the vehicle meets the ground.

[0062] In the present invention, the detection rules corresponding to different perception task types are pre-configured in the model and are obtained based on the correspondence between the perception task types and the detection rules.

[0063] The judgment parameters of the task objectives corresponding to each perception task type are determined according to the determined multiple detection rules and the image features of the image to be tested.

[0064] After the judgment parameters of the task objectives corresponding to each perception task type are obtained, the judgment parameters of the task objectives corresponding to each perception task type are input into the processing layer. Different perception task types also correspond to different judgment rules.

[0065] For example, if the perception task is whether a vehicle is changing lanes, and the judgment parameters are the vehicle's landing point and the position of the lane line, the judgment rule can be configured to determine the distance and angle between the vehicle's landing point and the lane line.

[0066] In the present invention, the processing layer determines the judgment rules corresponding to each perception task type based on multiple pre-stored perception task types, and makes judgments based on the determined multiple judgment rules and the judgment parameters corresponding to the multiple judgment rules to obtain the task attributes of the task target.

[0067] The driving task processing method provided by the present invention has a further method in which a detection layer and a processing layer are used to perform division of labor and cooperation on the image features of the image to be tested, thereby accelerating the processing speed.

[0068] In further explanation of the above method, within each perception task, some perception tasks may be related to each other. That is, when determining the task attributes of one perception task, the judgment parameters of another perception task are used. Therefore, it is necessary to obtain the correlation between the perception task types.

[0069] The image features of the image to be tested are input into the detection layer, which determines the detection rules corresponding to each perception task type based on multiple pre-stored perception task types, and determines the judgment parameters of the task targets corresponding to each perception task type based on the determined multiple detection rules and the image features of the image to be tested.

[0070] The judgment parameters of the task objectives corresponding to each perception task type are input into the processing layer. The processing layer determines the first and second types of perception task types based on the correlation between the perception task types. Here, the first type of perception task type is the perception task type with correlation, and the second type of perception task type is the perception task type without correlation. For example, if the task attributes of perception task A require the judgment parameters of perception task B during the judgment process, perception task A is the first type of perception task type, and perception task B is the second type of perception task type. In addition, other perception task types without correlation also belong to the second type of perception task type.

[0071] The corresponding judgment rules are determined based on the first and second perception task types. Specifically, since the first perception task type requires the judgment parameters of other perception tasks, it requires corresponding rules that can combine the judgment parameters of both perception task types. The second perception task type does not require the judgment parameters of other perception tasks, so it still uses rules that only judge based on its own judgment parameters.

[0072] Since there are related perception tasks, the judgment parameters of the task objectives corresponding to each perception task type are divided to obtain new judgment parameters. Here, the new judgment parameters can be a parameter set that combines two or more judgment parameters.

[0073] Judgment is performed based on the determined multiple judgment rules and the new judgment parameters corresponding to each of the multiple judgment rules to obtain the task attributes of the task target.

[0074] In further explanation, the distribution layer includes a partitioning layer and multiple sub-processing layers;

[0075] Accordingly, the division layer determines the first and second types of perception task types according to the correlation between the perception task types, determines the corresponding judgment rules according to the first and second types of perception task types, and divides the judgment parameters of the task targets corresponding to each perception task type to obtain new judgment parameters;

[0076] The sub-processing layer makes a judgment based on the assigned judgment rules and the corresponding new judgment parameters to obtain the task attributes of the task target.

[0077] In this regard, it should be noted that the judgment rules, perception task types and redivision of judgment parameters of this application all require accurate division and configuration of the division layers.

[0078] In this application, it is necessary to perform branch processing on the image to be tested. To this end, multiple sub-processing layers can be set up to execute the judgment parameters corresponding to each perception task in parallel and separately, so that each perception task can share a processing model, reduce the establishment of multiple models for different perception tasks, reduce the overall memory usage of the system, reduce the workload, and achieve the purpose of strong scalability and easy maintenance.

[0079] The driving task processing device provided by the present invention is described below. The driving task processing device described below and the driving task processing method described above can be referenced to each other.

[0080] Figure 2 A schematic diagram of the structure of a driving task processing device provided by the present invention is shown. Figure 2 , the device includes a collection module 21 and a processing module 22, wherein:

[0081] An acquisition module 21 is used to obtain image features of an image to be measured;

[0082] The processing module 22 is configured to input the image features of the image to be tested into a processing model to obtain task attributes of the task target in the image to be tested output by the processing model; the task attributes represent the perception results of the task target under different perception tasks;

[0083] The processing model is a model obtained through machine learning training that uses image features in image samples as input and the perception results of task targets in image samples under various perception tasks as output, and is used to determine task attributes in images.

[0084] The processing model includes a detection layer and a processing layer. The detection layer is used to determine the judgment parameters of the task target in the image to be tested according to the image features of the image to be tested; the processing layer is used to determine the task attributes of the task target according to the assigned judgment parameters.

[0085] In further description of the above device, the processing module is specifically used to:

[0086] Inputting the image features of the image to be tested into the detection layer, the detection layer determines the detection rules corresponding to each perception task type according to a plurality of pre-stored perception task types, and determines the judgment parameters of the task targets corresponding to each perception task type according to the determined plurality of detection rules and the image features of the image to be tested;

[0087] The judgment parameters of the task targets corresponding to each perception task type are input into the processing layer, and the processing layer determines the judgment rules corresponding to each perception task type based on multiple pre-stored perception task types. The judgment is performed based on the determined multiple judgment rules and the judgment parameters corresponding to the multiple judgment rules to obtain the task attributes of the task targets.

[0088] In a further description of the above apparatus, the acquisition module is further configured to: obtain correlations between perception task types;

[0089] Accordingly, the processing module is specifically configured to:

[0090] Inputting the image features of the image to be tested into the detection layer, the detection layer determines the detection rules corresponding to each perception task type according to a plurality of pre-stored perception task types, and determines the judgment parameters of the task targets corresponding to each perception task type according to the determined plurality of detection rules and the image features of the image to be tested;

[0091] The judgment parameters of the task targets corresponding to each perception task type are input into the processing layer, and the processing layer determines the first type of perception task type and the second type of perception task type according to the correlation between the perception task types, determines the corresponding judgment rules according to the first type of perception task type and the second type of perception task type, and divides the judgment parameters of the task targets corresponding to each perception task type to obtain new judgment parameters, and makes judgments based on the determined multiple judgment rules and the new judgment parameters corresponding to the multiple judgment rules to obtain the task attributes of the task targets.

[0092] In a further description of the above apparatus, the distribution layer includes a partitioning layer and a plurality of sub-processing layers;

[0093] Accordingly, the division layer determines the first and second types of perception task types according to the correlation between the perception task types, determines the corresponding judgment rules according to the first and second types of perception task types, and divides the judgment parameters of the task targets corresponding to each perception task type to obtain new judgment parameters;

[0094] The sub-processing layer performs judgment according to the assigned judgment rule and the corresponding new judgment parameter to obtain the task attributes of the task target.

[0095] Since the principles of the apparatus described in the embodiment of the present invention are the same as those of the method described in the above embodiment, more detailed explanations are omitted here.

[0096] It should be noted that, in the embodiment of the present invention, relevant functional modules can be implemented by a hardware processor.

[0097] The driving task processing device provided by the present invention realizes that each perception task shares a processing model by branching the test image, reducing the need to establish multiple models for different perception tasks, lowering the overall system memory usage, and reducing the workload, thereby achieving the goals of strong scalability and easy maintenance.

[0098] Figure 3 An example of a physical structure diagram of an electronic device is shown below. Figure 3 As shown, the electronic device may include: a processor 31, a communication interface 32, a memory 33, and a communication bus 34, wherein the processor 31, the communication interface 32, and the memory 33 communicate with each other via the communication bus 34. The processor 31 may call a computer program in the memory 33 to execute the steps of the driving task processing method, for example, including: obtaining image features of the image to be tested; inputting the image features of the image to be tested into a processing model to obtain task attributes of the task target in the image to be tested output by the processing model; the task attributes represent the perception results of the task target under different perception tasks;

[0099] The processing model is a model obtained through machine learning training that uses image features in image samples as input and the perception results of task targets in image samples under various perception tasks as output, and is used to determine task attributes in images.

[0100] The processing model includes a detection layer and a processing layer. The detection layer is used to determine the judgment parameters of the task target in the image to be tested according to the image features of the image to be tested; the processing layer is used to determine the task attributes of the task target according to the assigned judgment parameters.

[0101] In addition, the logic instructions in the above-mentioned memory 33 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0102] On the other hand, the present invention further provides a computer program product, comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions. When the program instructions are executed by a computer, the computer is capable of executing the driving task processing method provided by the above methods, the method comprising: obtaining image features of an image to be tested; inputting the image features of the image to be tested into a processing model to obtain task attributes of a task target in the image to be tested output by the processing model; the task attributes representing the perception results of the task target under different perception tasks;

[0103] The processing model is a model obtained through machine learning training that uses image features in image samples as input and the perception results of task targets in image samples under various perception tasks as output, and is used to determine task attributes in images.

[0104] The processing model includes a detection layer and a processing layer. The detection layer is used to determine the judgment parameters of the task target in the image to be tested according to the image features of the image to be tested; the processing layer is used to determine the task attributes of the task target according to the assigned judgment parameters.

[0105] On the other hand, an embodiment of the present application further provides a processor-readable storage medium storing a computer program, wherein the computer program is configured to cause the processor to execute the driving task processing method provided in each of the above embodiments, for example, including: obtaining image features of an image to be tested; inputting the image features of the image to be tested into a processing model to obtain task attributes of a task target in the image to be tested output by the processing model; the task attributes representing the perception results of the task target under different perception tasks;

[0106] The processing model is a model obtained through machine learning training that uses image features in image samples as input and the perception results of task targets in image samples under various perception tasks as output, and is used to determine task attributes in images.

[0107] The processing model includes a detection layer and a processing layer. The detection layer is used to determine the judgment parameters of the task target in the image to be tested according to the image features of the image to be tested; the processing layer is used to determine the task attributes of the task target according to the assigned judgment parameters.

[0108] The processor-readable storage medium can be any available medium or data storage device that can be accessed by the processor, including but not limited to magnetic storage (such as floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO)), optical storage (such as CDs, DVDs, BDs, HVDs, etc.), and semiconductor storage (such as ROMs, EPROMs, EEPROMs, non-volatile memories (NANDFLASH), solid-state drives (SSDs)), etc.

[0109] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0110] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiments.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A driving task processing method, characterized in that: include: Obtaining image features of the image to be tested; Inputting the image features of the image to be tested into a processing model to obtain task attributes of the task target in the image to be tested output by the processing model; the task attributes represent the perception results of the task target under different perception tasks; The processing model is a model obtained through machine learning training that uses image features in image samples as input and the perception results of task targets in image samples under various perception tasks as output, and is used to determine task attributes in images. The processing model includes a detection layer and a processing layer, wherein the detection layer is used to determine the judgment parameters of the task target in the image to be tested according to the image features of the image to be tested; and the processing layer is used to determine the task attributes of the task target according to the assigned judgment parameters; The method further includes: obtaining correlations between the types of sensing tasks; Accordingly, the image features of the image to be tested are input into the processing model to obtain the task attributes of the task target in the image to be tested output by the processing model, including: Inputting the image features of the image to be tested into the detection layer, the detection layer determines the detection rules corresponding to each perception task type according to a plurality of pre-stored perception task types, and determines the judgment parameters of the task targets corresponding to each perception task type according to the determined plurality of detection rules and the image features of the image to be tested; The judgment parameters of the task targets corresponding to each perception task type are input into the processing layer, and the processing layer determines the first category of perception task type and the second category of perception task type according to the correlation between the perception task types, determines the corresponding judgment rules according to the first category of perception task type and the second category of perception task type, and divides the judgment parameters of the task targets corresponding to each perception task type to obtain new judgment parameters, and makes judgments based on the determined multiple judgment rules and the new judgment parameters corresponding to the multiple judgment rules to obtain the task attributes of the task targets; wherein, the judgment rules of the first category of perception task type are rules that can be used to make judgments based on the judgment parameters of the two perception task types.

2. The driving task processing method according to claim 1, characterized in that: Inputting the image features of the image to be tested into the processing model to obtain the task attributes of the task target in the image to be tested output by the processing model includes: Inputting the image features of the image to be tested into the detection layer, the detection layer determines the detection rules corresponding to each perception task type according to a plurality of pre-stored perception task types, and determines the judgment parameters of the task targets corresponding to each perception task type according to the determined plurality of detection rules and the image features of the image to be tested; The judgment parameters of the task targets corresponding to each perception task type are input into the processing layer, and the processing layer determines the judgment rules corresponding to each perception task type based on multiple pre-stored perception task types. The judgment is performed based on the determined multiple judgment rules and the judgment parameters corresponding to the multiple judgment rules to obtain the task attributes of the task targets.

3. The driving task processing method according to claim 1, characterized in that: The distribution layer includes a partitioning layer and multiple sub-processing layers; Accordingly, the division layer determines the first and second types of perception task types according to the correlation between the perception task types, determines the corresponding judgment rules according to the first and second types of perception task types, and divides the judgment parameters of the task targets corresponding to each perception task type to obtain new judgment parameters; The sub-processing layer performs judgment according to the assigned judgment rule and the corresponding new judgment parameter to obtain the task attributes of the task target.

4. The driving task processing method according to any one of claims 1 to 3, characterized in that: The perception tasks include 2D / 3D target detection, obstacle detection, lane line detection, traffic sign detection, road marking detection, key point detection, drivable area segmentation, monocular depth estimation and optical flow estimation.

5. A driving task processing device, characterized in that: include: An acquisition module, used to obtain image features of the image to be tested; a processing module, configured to input the image features of the image to be tested into a processing model to obtain task attributes of the task target in the image to be tested output by the processing model; the task attributes represent the perception results of the task target under different perception tasks; The processing model is a model obtained through machine learning training that uses image features in image samples as input and the perception results of task targets in image samples under various perception tasks as output, and is used to determine task attributes in images. The processing model includes a detection layer and a processing layer, wherein the detection layer is used to determine the judgment parameters of the task target in the image to be tested according to the image features of the image to be tested; and the processing layer is used to determine the task attributes of the task target according to the assigned judgment parameters; The acquisition module is further used to: obtain the correlation between the perception task types; Accordingly, the processing module is specifically configured to: Inputting the image features of the image to be tested into the detection layer, the detection layer determines the detection rules corresponding to each perception task type according to a plurality of pre-stored perception task types, and determines the judgment parameters of the task targets corresponding to each perception task type according to the determined plurality of detection rules and the image features of the image to be tested; The judgment parameters of the task targets corresponding to each perception task type are input into the processing layer, and the processing layer determines the first category of perception task type and the second category of perception task type according to the correlation between the perception task types, determines the corresponding judgment rules according to the first category of perception task type and the second category of perception task type, and divides the judgment parameters of the task targets corresponding to each perception task type to obtain new judgment parameters, and makes judgments based on the determined multiple judgment rules and the new judgment parameters corresponding to the multiple judgment rules to obtain the task attributes of the task targets; wherein, the judgment rules of the first category of perception task type are rules that can be used to make judgments based on the judgment parameters of the two perception task types.

6. The driving task processing device according to claim 5, characterized in that: The processing module is specifically used for: Inputting the image features of the image to be tested into the detection layer, the detection layer determines the detection rules corresponding to each perception task type according to a plurality of pre-stored perception task types, and determines the judgment parameters of the task targets corresponding to each perception task type according to the determined plurality of detection rules and the image features of the image to be tested; The judgment parameters of the task targets corresponding to each perception task type are input into the processing layer, and the processing layer determines the judgment rules corresponding to each perception task type based on multiple pre-stored perception task types. The judgment is performed based on the determined multiple judgment rules and the judgment parameters corresponding to the multiple judgment rules to obtain the task attributes of the task targets.

7. An electronic device comprising a processor and a memory storing a computer program, characterized in that: When the processor executes the computer program, the steps of the driving task processing method according to any one of claims 1 to 4 are implemented.

8. A processor-readable storage medium, characterized in that: The processor-readable storage medium stores a computer program, and the computer program is used to enable the processor to execute the steps of the driving task processing method according to any one of claims 1 to 4.

Images