Driving scene perception detection method, device, electronic device and storage medium

Abstract

The present invention provides a driving scene perception detection method, device, electronic device and storage medium, the method comprising: obtaining scene features of driving scene data; inputting the scene features into a perception detection model, obtaining a perception result output by the perception detection model and an uncertainty corresponding to the perception result, wherein the perception result represents the result of whether a perception event exists in the driving scene; and the uncertainty represents the credibility of the perception event in the driving scene. The driving scene perception detection method, system, electronic device and storage medium provided by the present invention input the scene features into the perception detection model, and the perception detection model outputs the uncertainty of the perception result and the perception result synchronously, so that the credibility of the corresponding perception result can be synchronously provided when the perception result is used in subsequent work, and a better decision logic can be made by using the subsequent work.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a driving scene perception detection method, system, electronic device and storage medium. Background Art

[0002] There are many perception tasks in autonomous driving scenarios, such as 2D / 3D target detection, general obstacle detection, lane line detection, traffic sign detection, road sign detection, key point detection, drivable area segmentation, monocular depth estimation, optical flow estimation, etc.

[0003] Nowadays, the mainstream is to use supervised methods based on deep learning. However, the perception results output by the network model that analyzes the perception task often have a certain degree of uncertainty and different degrees of credibility. At the same time, there is a certain correlation between different perception tasks, but it cannot be expressed at present. Summary of the invention

[0004] In view of the problems existing in the prior art, the present invention provides a driving scene perception detection method, system, electronic device and storage medium.

[0005] In a first aspect, the present invention provides a driving scene perception detection method, comprising:

[0006] Acquire scene features of driving scene data;

[0007] Inputting the scene features into a perception detection model to obtain a perception result output by the perception detection model and an uncertainty corresponding to the perception result, wherein the perception result represents the result of whether a perception event exists in the driving scene; and the uncertainty represents the credibility of the perception event in the driving scene;

[0008] The perception detection model is a model obtained by machine learning training based on the scene features of the driving scene samples and the perception results of the samples as inputs, and is used to judge the perception results and uncertainty of the driving scene data;

[0009] The perception detection model includes a matching layer and a processing layer. The matching layer is used to determine the scene characteristics under different perception task types according to the scene characteristics of the driving scene data; the processing layer is used to determine the perception results and uncertainty under different perception task types according to the scene characteristics under different perception task types.

[0010] In one embodiment, the step of inputting the scene feature into a perception detection model to obtain a perception result output by the perception detection model and an uncertainty corresponding to the perception result includes:

[0011] Inputting scene features of the driving scene data into a matching layer, and the matching layer determines corresponding scene features according to different perception task types;

[0012] The scene features under different perception task types are input into the processing layer, and the processing layer determines the perception results and uncertainty under different perception task types according to the scene features under different perception task types.

[0013] In one embodiment, the method further comprises a step of acquiring a perception detection model, wherein the step of acquiring comprises:

[0014] Establishing an initial neural network model, configuring an initial loss value and a relationship between the loss value and the prediction result, and the uncertainty of the prediction result for the initial neural network model;

[0015] Inputting the scene features of the driving scene samples and the true value results of the samples into the initial neural network model, and determining the prediction results and the uncertainty corresponding to the prediction results according to the true value results of the samples, the initial loss value, and the relationship between the loss value and the prediction results and the uncertainty of the prediction results;

[0016] A calibration loss value is determined according to the prediction result and the true value result, the initial loss value is calibrated according to the calibration loss value, and the optimized neural network model is determined as the perception detection model when the loss value meets the preset conditions.

[0017] In one embodiment, the relationship between the loss value, the prediction result, and the uncertainty of the prediction result includes:

[0018]

[0019] Among them, L loc is the loss value, μ * is the true value result of the sample, μ is the predicted result, Σ is the covariance matrix of uncertainty under multi-sensory tasks, and T is the matrix transpose.

[0020] In a second aspect, the present invention provides a driving scene perception detection device, comprising:

[0021] An acquisition module, used to acquire scene features of driving scene data;

[0022] A processing module, used for inputting the scene features into a perception detection model to obtain a perception result output by the perception detection model and an uncertainty corresponding to the perception result, wherein the perception result represents the result of whether a perception event exists in the driving scene; and the uncertainty represents the credibility of the perception event in the driving scene;

[0023] The perception detection model is a model obtained by machine learning training based on the scene features of the driving scene samples and the perception results of the samples as inputs, and is used to judge the perception results and uncertainty of the driving scene data;

[0024] The perception detection model includes a matching layer and a processing layer. The matching layer is used to determine the scene characteristics under different perception task types according to the scene characteristics of the driving scene data; the processing layer is used to determine the perception results and uncertainty under different perception task types according to the scene characteristics under different perception task types.

[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 scene perception detection 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 scene perception detection method described in the first aspect.

[0027] The driving scene perception detection method, system, electronic device and storage medium provided by the present invention, by inputting scene features into a perception detection model, the perception detection model synchronously outputs the uncertainty of the perception result and the perception result, can synchronously provide the credibility of the corresponding perception result when the perception result is used in subsequent work, and can make better decision logic by utilizing the subsequent work. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces 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 creative work.

[0029] Figure 1 It is a flow chart of the driving scene perception detection method provided by the present invention;

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

[0031] Figure 3 is a schematic diagram of the structure of an electronic device provided by the present invention; DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] Combine the following Figure 1-Figure 3 The present invention describes a driving scene perception detection method, system, electronic device and storage medium.

[0034] Figure 1 A schematic diagram of a driving scene perception detection method of the present invention is shown in FIG. Figure 1 , the method comprising:

[0035] 11. Obtain scene features of driving scene data;

[0036] 12. Input the scene features into the perception detection model to obtain the perception results output by the perception detection model and the uncertainty corresponding to the perception results. The perception results represent the results of whether there is a perception event in the driving scene; the uncertainty represents the credibility of the perception event in the driving scene. Among them, the perception detection model is a model obtained by machine learning training based on the scene features of the driving scene samples and the perception results of the samples as inputs, and is used to judge the perception results and uncertainty of the driving scene data;

[0037] The perception detection model includes a matching layer and a processing layer. The matching layer is used to determine the scene characteristics under different perception task types according to the scene characteristics of the driving scene data; the processing layer is used to determine the perception results and uncertainty under different perception task types according to the scene characteristics under different perception task types.

[0038] Regarding steps 11 and 12, it should be noted that in the present invention, there are many perception tasks in the autonomous driving scenario, such as 2D / 3D target detection, general obstacle detection, lane line detection, traffic sign detection, road surface sign detection, key point detection, drivable area segmentation, monocular depth estimation, optical flow estimation, etc.

[0039] Different perception tasks correspond to different perception results. For example, 2D object detection corresponds to the perception result of marking the object with a 2D box on the screen, and the data based on which the 2D box is generated is the perception result.

[0040] However, the perception results output by the network model for analyzing the perception task often have a certain degree of uncertainty and different degrees of credibility. Therefore, in the present invention, the perception detection model for analyzing the perception task is a model obtained by machine learning training based on the scene features of the driving scene samples and the perception results of the samples as inputs, and is used to judge the perception results and uncertainty of the driving scene data.

[0041] In the present invention, the data used to analyze the perception task is the scene features of the driving scene data. To this end, it is necessary to obtain the driving scene data and extract the scene features. The scene features are then input into the perception detection model. For example, for perception tasks such as 2D / 3D target detection, general obstacle detection, lane line detection, traffic sign detection, and road surface marking detection, the driving scene data may be a video image of the driving vehicle's surroundings.

[0042] It should be noted that users will generate a large amount of driving scene data during the autonomous driving process, and these driving scene data will be manually annotated with corresponding perception results in advance. These driving scene data are used as samples, and the corresponding scene features are extracted from the samples. The scene features and perception results are input into the initial neural network model for training to obtain the perception detection model.

[0043] In the present invention, the perception detection model includes a matching layer and a processing layer. The matching layer is used to determine the scene characteristics under different perception task types according to the scene characteristics of the driving scene data; the processing layer is used to determine the perception results and uncertainty under different perception task types according to the scene characteristics under different perception task types.

[0044] In this regard, it should be noted that since there are many perception tasks in the autonomous driving scene, different perception tasks may be based on different scene features. For example, lane line detection is based on the characteristics of the lane line; target detection is based on the characteristics of the target object.

[0045] To this end, the matching layer needs to determine the scene features under different perception task types based on the scene features of the driving scene data. Then the processing layer needs to determine the perception results and uncertainty under different perception task types based on the scene features under different perception task types.

[0046] The driving scene perception detection method provided by the present invention, by inputting scene features into a perception detection model, and the perception detection model synchronously outputs the uncertainty of the perception result and the perception result, can synchronously provide the credibility of the corresponding perception result when the perception result is used in subsequent work, and can make better decision logic by utilizing the subsequent work.

[0047] In the further description of the above method, the processing of inputting the scene features into the perception detection model, obtaining the perception results output by the perception detection model and the uncertainty corresponding to the perception results is mainly explained, as follows:

[0048] The scene features of the driving scene data are input into the matching layer, and the matching layer determines the corresponding scene features according to different perception task types;

[0049] The scene features under different perception task types are input into the processing layer, and the processing layer determines the perception results and uncertainty under different perception task types according to the scene features under different perception task types.

[0050] In this regard, it should be noted that in the present invention, since there are many perception tasks in the autonomous driving scene, different perception tasks may be based on different scene features. For example, lane line detection is based on the characteristics of the lane line; target detection is based on the characteristics of the target object.

[0051] To this end, the matching layer needs to determine the scene features under different perception task types based on the scene features of the driving scene data. Then the processing layer needs to determine the perception results and uncertainty under different perception task types based on the scene features under different perception task types.

[0052] A further method of the present invention uses a matching layer and a processing layer to divide and cooperate in processing scene features corresponding to different perception tasks, which can speed up the processing speed.

[0053] In the further description of the above method, the acquisition process of the perception detection model is mainly explained, as follows:

[0054] Establish an initial neural network model, configure an initial loss value and a relationship between the loss value and the prediction result, and the uncertainty of the prediction result for the initial neural network model;

[0055] Input the scene features of the driving scene samples and the true value results of the samples into the initial neural network model, and determine the prediction results and the uncertainty corresponding to the prediction results according to the true value results of the samples, the initial loss value, and the relationship between the loss value and the prediction results and the uncertainty of the prediction results;

[0056] A calibration loss value is determined according to the prediction result and the true value result, the initial loss value is calibrated according to the calibration loss value, and the optimized neural network model is determined as the perception detection model when the loss value meets the preset conditions.

[0057] In this regard, it should be noted that in the process of model training, the neural network model in the initial state must first be constructed, and the initial loss value must be configured for the neural network model. The loss value is the value obtained by comparing the true value with the predicted value; the neural network model is also configured with a relationship between the loss value and the predicted result, and the uncertainty of the predicted result. This relationship characterizes the association between the loss value, the perception result (the true value of the training sample), the predicted result, and the uncertainty of the predicted result. Here, since the perception result belongs to the true value, this association is called the relationship between the loss value and the predicted result, and the uncertainty of the predicted result.

[0058] The scene features of the driving scene samples and the true value results of the samples are input into the initial neural network model, and then the sample data are analyzed one by one. The prediction results and the uncertainty corresponding to the prediction results are determined according to the relationship between the true value results of the samples, the initial loss value, the loss value and the prediction results, and the uncertainty of the prediction results.

[0059] The calculation is completed based on the comparison between the predicted result and the true value result, and a new loss value can be determined, which is used as the calibration loss value. Then the initial loss value is calibrated according to the calibration loss value, and the calibrated loss value is applied to the next calculation process until the loss value after multiple calibrations can meet the preset conditions (such as sufficient convergence). At this time, the optimized neural network model obtained can be used as the perception detection model of this method.

[0060] In addition, further explanation is required. The relationship between the loss value configured in the present invention and the prediction result and the uncertainty of the prediction result includes:

[0061]

[0062] Among them, L loc is the loss value, μ * is the true value result of the sample, μ is the predicted result, Σ is the covariance matrix of uncertainty under multi-sensory tasks, and T is the matrix transpose. Among them, the covariance matrix can not only reflect the uncertainty of each true value result, but also characterize the correlation between different true value results.

[0063] A further method of the present invention, through detection suitable for multi-perception tasks, establishes a relationship between the loss value and the prediction result, and the uncertainty of the prediction result, and can train a perception detection model that outputs the perception result and the uncertainty. It can also provide the corresponding credibility of the perception result when the perception result is used in subsequent work, and make better decision logic by utilizing the subsequent work.

[0064] The driving scene perception detection device provided by the present invention is described below. The driving scene perception detection device described below and the driving scene perception detection method described above can be referenced to each other.

[0065] Figure 2 The schematic diagram of the structure of a driving scene perception detection device provided by the present invention is shown. Referring to the figure, the device includes an acquisition module 21 and a processing module 22, wherein:

[0066] An acquisition module 21 is used to acquire scene features of driving scene data;

[0067] The processing module 22 is used to input the scene features into the perception detection model to obtain the perception result output by the perception detection model and the uncertainty corresponding to the perception result, wherein the perception result represents the result of whether the perception event exists in the driving scene; and the uncertainty represents the credibility of the perception event in the driving scene;

[0068] The perception detection model is a model that uses the scene features of the driving scene samples and the perception results of the samples as inputs and is obtained through machine learning training to judge the perception results and uncertainty of the driving scene data;

[0069] The perception detection model includes a matching layer and a processing layer. The matching layer is used to determine the scene characteristics under different perception task types according to the scene characteristics of the driving scene data; the processing layer is used to determine the perception results and uncertainty under different perception task types according to the scene characteristics under different perception task types.

[0070] In a further description of the above device, the processing module is specifically used for:

[0071] Inputting scene features of the driving scene data into a matching layer, and the matching layer determines corresponding scene features according to different perception task types;

[0072] The scene features under different perception task types are input into the processing layer, and the processing layer determines the perception results and uncertainty under different perception task types according to the scene features under different perception task types.

[0073] In a further description of the above device, the device also includes a building block for:

[0074] Establishing an initial neural network model, configuring an initial loss value and a relationship between the loss value and the prediction result, and the uncertainty of the prediction result for the initial neural network model;

[0075] Inputting the scene features of the driving scene samples and the true value results of the samples into the initial neural network model, and determining the prediction results and the uncertainty corresponding to the prediction results according to the true value results of the samples, the initial loss value, and the relationship between the loss value and the prediction results and the uncertainty of the prediction results;

[0076] A calibration loss value is determined according to the prediction result and the true value result, the initial loss value is calibrated according to the calibration loss value, and the optimized neural network model is determined as the perception detection model when the loss value meets the preset conditions.

[0077] In a further description of the above device, the relationship between the loss value and the prediction result and the uncertainty of the prediction result includes:

[0078]

[0079] Among them, L loc is the loss value, μ * is the true value result of the sample, μ is the predicted result, Σ is the covariance matrix of uncertainty under multi-sensory tasks, and T is the matrix transpose.

[0080] Since the principle of the device described in the embodiment of the present invention is the same as that of the method described in the above embodiment, a more detailed explanation will not be repeated here.

[0081] It should be noted that, in the embodiment of the present invention, the relevant functional modules can be implemented by a hardware processor.

[0082] The driving scene perception detection device provided by the present invention inputs scene features into a perception detection model, and the perception detection model synchronously outputs the uncertainty of the perception result and the perception result. It can synchronously provide the credibility of the corresponding perception result when the perception result is used in subsequent work, and make better decision logic by utilizing the subsequent work.

[0083] Figure 3 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 3As 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 through the communication bus 34. The processor 31 may call a computer program in the memory 33 to execute the steps of the driving scene perception detection method, for example, including: obtaining scene features of driving scene data; inputting the scene features into a perception detection model to obtain a perception result output by the perception detection model and an uncertainty corresponding to the perception result, wherein the perception result represents the result of whether a perception event exists in the driving scene; the uncertainty represents the credibility of the perception event in the driving scene;

[0084] The perception detection model is a model that uses the scene features of the driving scene samples and the perception results of the samples as inputs and is obtained through machine learning training to judge the perception results and uncertainty of the driving scene data;

[0085] The perception detection model includes a matching layer and a processing layer. The matching layer is used to determine the scene characteristics under different perception task types according to the scene characteristics of the driving scene data; the processing layer is used to determine the perception results and uncertainty under different perception task types according to the scene characteristics under different perception task types.

[0086] 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 it is sold or used as an independent product. Based on such an 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, and the computer software product is stored in a storage medium, including a number of instructions to enable 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 and other media that can store program codes.

[0087] On the other hand, the present invention further provides a computer program product, the 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 can execute the driving scene perception detection method provided by the above methods, the method comprising: acquiring scene features of driving scene data; inputting the scene features into a perception detection model, obtaining a perception result output by the perception detection model and an uncertainty corresponding to the perception result, the perception result representing the result of whether a perception event exists in the driving scene; the uncertainty representing the credibility of the perception event in the driving scene;

[0088] The perception detection model is a model that uses the scene features of the driving scene samples and the perception results of the samples as inputs and is obtained through machine learning training to judge the perception results and uncertainty of the driving scene data;

[0089] The perception detection model includes a matching layer and a processing layer. The matching layer is used to determine the scene characteristics under different perception task types according to the scene characteristics of the driving scene data; the processing layer is used to determine the perception results and uncertainty under different perception task types according to the scene characteristics under different perception task types.

[0090] On the other hand, an embodiment of the present application further provides a processor-readable storage medium, wherein the processor-readable storage medium stores a computer program, wherein the computer program is used to enable the processor to execute the driving scene perception detection method provided by the above embodiments, for example, including: acquiring scene features of driving scene data; inputting the scene features into a perception detection model to obtain a perception result output by the perception detection model and an uncertainty corresponding to the perception result, wherein the perception result represents the result of whether a perception event exists in the driving scene; and the uncertainty represents the credibility of the perception event in the driving scene;

[0091] The perception detection model is a model that uses the scene features of the driving scene samples and the perception results of the samples as inputs and is obtained through machine learning training to judge the perception results and uncertainty of the driving scene data;

[0092] The perception detection model includes a matching layer and a processing layer. The matching layer is used to determine the scene characteristics under different perception task types according to the scene characteristics of the driving scene data; the processing layer is used to determine the perception results and uncertainty under different perception task types according to the scene characteristics under different perception task types.

[0093] 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), etc.), optical storage (such as CD, DVD, BD, HVD, etc.), and semiconductor storage (such as ROM, EPROM, EEPROM, non-volatile memory (NANDFLASH), solid-state drive (SSD)), etc.

[0094] The device embodiments described above are merely illustrative, 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 scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0095] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for 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 some parts of the embodiments.

[0096] 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 embodiments of the present invention.

Claims

1. A driving scene perception detection method, characterized in that: include: Acquire scene features of driving scene data; Inputting the scene feature into a perception detection model to obtain a perception result output by the perception detection model and an uncertainty corresponding to the perception result, wherein the perception result represents the result of whether a perception event exists in the driving scene; The uncertainty represents the credibility of the perceived event in the driving scenario; The perception detection model is a model obtained by machine learning training based on the scene features of the driving scene samples and the perception results of the samples as inputs, and is used to judge the perception results and uncertainty of the driving scene data; The perception detection model includes a matching layer and a processing layer, wherein the matching layer is used to determine scene features under different perception task types according to scene features of driving scene data; The processing layer is used to determine the perception results and uncertainty under different perception task types according to the scene characteristics under different perception task types.

2. The driving scene perception detection method according to claim 1, characterized in that: The step of inputting the scene feature into the perception detection model to obtain the perception result output by the perception detection model and the uncertainty corresponding to the perception result includes: Inputting scene features of the driving scene data into a matching layer, and the matching layer determines corresponding scene features according to different perception task types; The scene features under different perception task types are input into the processing layer, and the processing layer determines the perception results and uncertainty under different perception task types according to the scene features under different perception task types.

3. The driving scene perception detection method according to claim 1, characterized in that: The method further comprises a step of acquiring a perception detection model, wherein the step of acquiring comprises: Establishing an initial neural network model, configuring an initial loss value and a relationship between the loss value and the prediction result, and the uncertainty of the prediction result for the initial neural network model; Inputting the scene features of the driving scene samples and the true value results of the samples into the initial neural network model, and determining the prediction results and the uncertainty corresponding to the prediction results according to the true value results of the samples, the initial loss value, and the relationship between the loss value and the prediction results and the uncertainty of the prediction results; A calibration loss value is determined according to the prediction result and the true value result, the initial loss value is calibrated according to the calibration loss value, and the optimized neural network model is determined as the perception detection model when the loss value meets the preset conditions.

4. The driving scene perception detection method according to claim 3, characterized in that: The relationship between the loss value, the prediction result, and the uncertainty of the prediction result includes: Among them, L loc is the loss value, μ * is the true value result of the sample, μ is the predicted result, Σ is the covariance matrix of uncertainty under multi-sensory tasks, and T represents matrix transpose.

5. A driving scene perception detection device, characterized in that: include: An acquisition module, used to acquire scene features of driving scene data; A processing module, used for inputting the scene features into a perception detection model to obtain a perception result output by the perception detection model and an uncertainty corresponding to the perception result, wherein the perception result represents the result of whether there is a perception event in the driving scene; The uncertainty represents the credibility of the perceived event in the driving scenario; The perception detection model is a model obtained by machine learning training based on the scene features of the driving scene samples and the perception results of the samples as inputs, and is used to judge the perception results and uncertainty of the driving scene data; The perception detection model includes a matching layer and a processing layer, wherein the matching layer is used to determine scene features under different perception task types according to scene features of driving scene data; The processing layer is used to determine the perception results and uncertainty under different perception task types according to the scene characteristics under different perception task types.

6. The driving scene perception detection device according to claim 5, characterized in that: The processing module is specifically used for: Inputting scene features of the driving scene data into a matching layer, and the matching layer determines corresponding scene features according to different perception task types; The scene features under different perception task types are input into the processing layer, and the processing layer determines the perception results and uncertainty under different perception task types according to the scene features under different perception task types.

7. The driving scene perception detection device according to claim 5, characterized in that: The apparatus further comprises a building block for: Establishing an initial neural network model, configuring an initial loss value and a relationship between the loss value and the prediction result, and the uncertainty of the prediction result for the initial neural network model; Inputting the scene features of the driving scene samples and the true value results of the samples into the initial neural network model, and determining the prediction results and the uncertainty corresponding to the prediction results according to the true value results of the samples, the initial loss value, and the relationship between the loss value and the prediction results and the uncertainty of the prediction results; A calibration loss value is determined according to the prediction result and the true value result, the initial loss value is calibrated according to the calibration loss value, and the optimized neural network model is determined as the perception detection model when the loss value meets the preset conditions.

8. The driving scene perception detection device according to claim 7, characterized in that: The relationship between the loss value, the prediction result, and the uncertainty of the prediction result includes: Among them, L loc is the loss value, μ * is the true value result of the sample, μ is the predicted result, Σ is the covariance matrix of uncertainty under multi-sensory tasks, and T represents matrix transpose.

9. 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 scene perception detection method according to any one of claims 1 to 4 are implemented.

10. 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 scene perception detection method according to any one of claims 1 to 4.

Images