Method and device for identifying folding state of rearview mirror and electronic equipment

Abstract

The invention relates to a method and device for recognizing the folding state of a rearview mirror and electronic equipment, and belongs to the technical field of aided driving, and the method comprises the steps: obtaining and preprocessing image data to obtain training set data, verification set data and test set data, and training, verifying and testing the convolutional neural network model according to the training set data, the verification set data and the test set data to obtain a folding state determination model, deploying the folding state determination model in the vehicle-mounted auxiliary system, and when the vehicle-mounted auxiliary system identifies the current folding state of the rearview mirror, determining the folding state of the rearview mirror. Inputting the image data acquired in real time into the folding state determination model to obtain the folding state of the rearview mirror; according to the method, the folding state of the rearview mirror can be accurately recognized in various environments by combining the image data acquired by the left rearview mirror camera and the right rearview mirror camera of the vehicle in various environments and the panoramic aerial view of the vehicle body based on an image recognition technology of deep learning.

Description

Technical Field

[0001] This application relates to the field of assisted driving technology, and particularly to a method, device, and electronic device for identifying the folded state of rearview mirrors. Background Art

[0002] In the field of autonomous driving, in order to eliminate blind spots, an ideal solution for vehicles is to use a panoramic parking system with four cameras. As mentioned in the patent application No. WOCN14087554, Method and Device for Image Processing for Automobiles, Method for Generating an Omnidirectional View Image of an Automobile, and Automobile Omnidirectional View System, the positions of the four cameras are respectively located on the front grille of the vehicle body, the trunk lid, and the left and right exterior rearview mirrors. If the parking system is to be used normally, the left and right exterior rearview mirrors should be unfolded normally. If the left and right exterior rearview mirrors are in the folded state, starting the autonomous parking system at this time may cause parking failure and abnormal obstacle recognition, ultimately leading to system safety risks. Therefore, in the design of the autonomous parking assistance system, it is necessary to monitor the folded state of the vehicle's rearview mirrors.

[0003] Existing methods for detecting the folded state of rearview mirrors generally rely on angle sensors or folding motor feedback on the rearview mirrors. The folded angle of the rearview mirror is detected by the angle sensor or the current feedback of the rearview mirror folding motor is used for judgment to determine the folded state of the rearview mirror. However, the angle sensor and folding motor feedback are easily affected by environmental factors such as weather, and it is easy to cause inaccurate recognition of the folded state under harsh conditions. Summary of the Invention

[0004] To this end, the present invention provides a method, device, and electronic device for identifying the folded state of rearview mirrors to solve the problem of inaccurate recognition of the folded state caused by the susceptibility of angle sensors and folding motor feedback to environmental factors such as weather in the prior art.

[0005] In a first aspect, a method for identifying the folded state of rearview mirrors is provided. The method includes:

[0006] Obtaining and preprocessing image data to obtain training set data, validation set data, and test set data; the image data includes image data collected by left and right rearview mirror cameras and a bird's-eye view of the vehicle body in a preset scenario; the preset scenario includes an environmental scenario formed by combining a preset weather, a preset vehicle state, and a preset time point; the preset weather includes sunny, rainy, snowy, and foggy days; the preset vehicle state includes stationary, slow driving, and normal driving; the preset time point is each hour of the 24 hours of a day, and each hour is a preset time point;

[0007] Training, validating, and testing a convolutional neural network model based on the training set data, validation set data, and test set data to obtain a folded state determination model;

[0008] Deploy the folding state determination model to the vehicle-mounted auxiliary system;

[0009] When the vehicle-mounted auxiliary system recognizes the current folding state of the rearview mirror, input the acquired image data in real time into the folding state determination model to obtain the folding state of the rearview mirror.

[0010] Further, the acquired image data will be preprocessed to obtain training set data, validation set data, and test set data, including:

[0011] Acquire the image data collected by the left and right rearview mirror cameras and the bird's-eye view of the vehicle body surround view in a preset scenario as the image data;

[0012] Perform denoising processing on all images in the image data to obtain target image data;

[0013] Label all images in the target image data one by one and divide them into training set data, validation set data, and test set data; the training set data accounts for a first preset proportion of the target image data; the validation set data accounts for a second preset proportion of the target image data; the test set data accounts for a third preset proportion of the target image data.

[0014] Further, the performing denoising processing on all images in the image data to obtain target image data includes:

[0015] Perform a filtering operation on all images in the image data through a filtering algorithm to reduce noise interference caused by weather and light;

[0016] Adjust the brightness and contrast of all images in the image data to ensure consistency of all images during model training.

[0017] Further, the labeling all images in the target image data one by one includes:

[0018] Label all images in the target image data one by one through a labeling tool; the labeled folding states include fully folded, partially folded, and unfolded.

[0019] Further, the training, validating, and testing the convolutional neural network model based on the training set data, validation set data, and test set data to obtain the folding state determination model includes:

[0020] Training step: Train the convolutional neural network model based on the training set data to obtain a training result;

[0021] Hyperparameter adjustment step: Adjust the hyperparameters of the convolutional neural network model based on the training result;

[0022] Verification step: verifying the adjusted convolutional neural network model based on the verification set data; the model verification is based on the verification of loss and the recognition accuracy of the folding state;

[0023] Testing step: testing the adjusted convolutional neural network model according to the test set to obtain the recognition accuracy of the folding state;

[0024] If the recognition accuracy is less than the preset accuracy, then the training step, the hyperparameter adjustment step, the verification step, and the testing step are sequentially executed again;

[0025] If the recognition accuracy is greater than or equal to the preset accuracy, the model training ends to obtain a folding state determination model.

[0026] Further, after inputting the image data obtained in real time into the folding state determination model, obtaining the folding state of the rearview mirror includes:

[0027] Inputting the image data collected by the left and right rearview mirror cameras in real time into the folding state determination model to obtain the real-time folding state of the rearview mirror, or,

[0028] If there is occlusion in the image data collected by the left and right rearview mirror cameras in real time, inputting the real-time obtained bird's-eye view of the vehicle body around into the folding state determination model to obtain the real-time folding state of the rearview mirror.

[0029] Further, the method further includes:

[0030] If the real-time folding state of the rearview mirror is empty, determining the real-time folding state of the rearview mirror through the angle sensor of the rearview mirror.

[0031] Further, the method further includes:

[0032] If the real-time folding state of the rearview mirror is empty, determining the real-time folding state of the rearview mirror through the feedback of the folding motor of the rearview mirror.

[0033] In a second aspect, a device for recognizing the folding state of a rearview mirror is provided, and the device includes:

[0034] An acquisition module, configured to acquire and preprocess image data to obtain training set data, verification set data, and test set data; the image data includes the image data collected by the left and right rearview mirror cameras in a preset scenario, and the bird's-eye view of the vehicle body around; the preset scenario includes an environmental scenario formed by combining a preset weather, a preset vehicle state, and a preset time point; the preset weather includes sunny, rainy, snowy, and foggy; the preset vehicle state includes stationary, slow driving, and normal driving; the preset time point is each hour in 24 hours of a day as a preset time point;

[0035] A model training module, configured to train, validate, and test a convolutional neural network model based on the training set data, validation set data, and test set data to obtain a folding state determination model;

[0036] A deployment module, configured to deploy the folding state determination model to an in-vehicle auxiliary system;

[0037] A folding state determination module, configured to input the real-time acquired image data into the folding state determination model to obtain the rearview mirror folding state when the in-vehicle auxiliary system recognizes the current rearview mirror folding state.

[0038] In a third aspect, an electronic device is provided, including:

[0039] At least one processor; and

[0040] A memory communicatively connected to the at least one processor; wherein,

[0041] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method for recognizing the rearview mirror folding state as described in any one of the above.

[0042] The present invention adopts the above technical solutions and at least has the following beneficial effects:

[0043] A method, device, and electronic device for recognizing the rearview mirror folding state are provided. The image data is acquired and preprocessed to obtain training set data, validation set data, and test set data. The image data includes image data collected by the left and right rearview mirror cameras and the body surround bird's-eye view under various environments. A folding state determination model is obtained by training, validating, and testing a convolutional neural network model based on the training set data, validation set data, and test set data. The folding state determination model is deployed to an in-vehicle auxiliary system. When the in-vehicle auxiliary system recognizes the current rearview mirror folding state, the real-time acquired image data is input into the folding state determination model to obtain the rearview mirror folding state. This method is based on the image recognition technology of deep learning, combines the image data collected by the left and right rearview mirror cameras and the body surround bird's-eye view of the vehicle under various environments, and can accurately recognize the folding state of the rearview mirror in various environments.

[0044] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit this application. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following briefly introduces the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0046] Figure 1 is a flowchart of a method for identifying the folded state of a rearview mirror shown in an exemplary embodiment of the present invention;

[0047] Figure 2 is a schematic block diagram of a device for identifying the folded state of a rearview mirror shown in an exemplary embodiment of the present invention;

[0048] Figure 3 is a schematic block diagram of an electronic device shown in an exemplary embodiment of the present invention. Detailed implementation manners

[0049] In order to make the purpose, technical solutions and advantages of the present application clearer, the following further details the present application in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0050] In the field of autonomous driving, in order to eliminate blind spots, an ideal solution for vehicles is to use a panoramic parking system with four cameras. As mentioned in the patent application No. WOCN14087554, Method and Device for Image Processing for Automobile, Method for Generating Surround View Image of Automobile and Automobile Surround View System, the positions of the four cameras are respectively located on the front grille, trunk lid and left and right exterior rearview mirrors of the vehicle body. If the parking system needs to be used normally, the left and right exterior rearview mirrors should be unfolded normally. If the left and right exterior rearview mirrors are in the folded state, starting the automatic parking system at this time may lead to parking failure and abnormal obstacle recognition, ultimately resulting in system safety risks. Therefore, in the design of the automatic parking assistance system, it is necessary to monitor the folded state of the vehicle's rearview mirrors. Currently, the common methods for detecting the folded state of rearview mirrors mainly include: 1. Angle sensor: The folded angle of the rearview mirror is detected by the sensor. However, the sensor is easily interfered by environmental factors such as weather and cannot guarantee accuracy under harsh conditions. 2. Motor feedback: The judgment is made through the angle feedback of the rearview mirror motor. For rearview mirrors using motors, the rotation angle will be transmitted to the ECU through the CAN bus, and then it is judged whether it is folded. However, this solution not only increases the manufacturing cost and system complexity, but is also easily interfered by environmental factors such as weather.

[0051] Embodiments of the present application provide a method, device, and electronic device for identifying the folding state of a rearview mirror. Based on deep learning image recognition technology and combined with images from the vehicle's four-way surround view cameras, the folding state of the rearview mirror can be accurately identified, with the following beneficial effects: High-precision recognition: By using the ResNet model and combining image processing technology, high-precision recognition of the rearview mirror folding state can be achieved in various complex environments, with an identification accuracy rate of over 99%. Reducing hardware costs: Compared with traditional sensor detection solutions, the present invention only needs to utilize the existing cameras of the vehicle, avoiding the dependence on additional sensors and reducing manufacturing and maintenance costs. Adapt to various environments: By collecting and training data under different weather and light conditions, the system can adapt to the usage requirements in harsh environments and ensure the stability of recognition. Enhancing system security: Real-time detection of the folding state of the rearview mirror avoids parking failures or obstacle recognition errors caused by the rearview mirror not being unfolded, thereby improving the safety during the parking process. Backup plan: In special cases, the bird's-eye view generated by the four-way surround view cameras can be used to replace the rearview mirror camera image for status judgment, ensuring the robustness and reliability of the system.

[0052] The methods and devices in the present application will be described below through specific embodiments.

[0053] Please refer to Figure 1 , Figure 1 which is a flowchart of a method for identifying the folding state of a rearview mirror shown in an exemplary embodiment of the present invention. Refer to Figure 1 , the method includes:

[0054] Step S11: Obtain and preprocess the image data to obtain training set data, validation set data, and test set data;

[0055] Step S12: Train, validate, and test the convolutional neural network model based on the training set data, validation set data, and test set data to obtain a folding state determination model;

[0056] Step S13: Deploy the folding state determination model to the in-vehicle assistance system;

[0057] Step S14: When the in-vehicle assistance system identifies the current folding state of the rearview mirror, input the real-time acquired image data into the folding state determination model to obtain the folding state of the rearview mirror.

[0058] It should be noted that the technical solution provided in this embodiment can be added to the existing in-vehicle assistance system in the form of a small program in specific practice, or it can also be in the form of an independent application program, providing an interface externally to complete the function of identifying the folding state of the rearview mirror; applicable scenarios include but are not limited to: intelligent driving, autonomous parking assistance systems, vehicle autonomous parking, and parking safety judgment.

[0059] Specifically, the image data includes the image data collected by the left and right rearview mirror cameras and the bird's-eye view of the vehicle body in a preset scenario; the preset scenario includes an environmental scenario formed by combining a preset weather, a preset vehicle state, and a preset time point; the preset weather includes sunny, rainy, snowy, and foggy days; the preset vehicle state includes stationary, slow driving, and normal driving; the preset time point is each hour of the 24 hours of a day, and each hour is a preset time point.

[0060] It can be understood that the method provided in this embodiment, based on the image recognition technology of deep learning, combines the image data collected by the left and right rearview mirror cameras of the vehicle in various environments and the bird's-eye view of the vehicle body, and can accurately identify the folding state of the rearview mirror in various environments.

[0061] In specific practice, step S11, "obtain and preprocess the image data to obtain training set data, validation set data, and test set data", includes: obtaining the image data collected by the left and right rearview mirror cameras and the bird's-eye view of the vehicle body in a preset scenario as the image data; performing denoising processing on all the images in the image data to obtain target image data; labeling each of the images in the target image data one by one and dividing them into training set data, validation set data, and test set data; the training set data accounts for a first preset ratio of the target image data; the validation set data accounts for a second preset ratio of the target image data; the test set data accounts for a third preset ratio of the target image data.

[0062] It should be noted that the bird's-eye view of the vehicle body is obtained by merging the fisheye images collected by the four cameras in the front, rear, left, and right of the vehicle body through the existing technology; the first preset ratio, the second preset ratio, and the third preset ratio can be set to 70%, 15%, and 15% in sequence, or can be set again according to the results of model training, validation, and testing.

[0063] Specifically, performing denoising processing on all the images in the image data to obtain target image data includes: performing a filtering operation on all the images in the image data through a filtering algorithm to reduce the noise interference caused by weather and light; adjusting the brightness and contrast of all the images in the image data to ensure the consistency of all the images during model training.

[0064] Specifically, labeling each of the images in the target image data one by one includes: labeling each of the images in the target image data one by one through a labeling tool; the labeled folding states include fully folded, partially folded, and unfolded.

[0065] In specific practice, step S12, "obtaining a folding state determination model after training, validating, and testing a convolutional neural network model based on training set data, validation set data, and test set data" includes: Training step: training the convolutional neural network model based on the training set data to obtain a training result; Hyperparameter adjustment step: adjusting the hyperparameters of the convolutional neural network model based on the training result; Validation step: validating the adjusted convolutional neural network model based on the validation set data; The model validation is based on the loss and the recognition accuracy of the folding state; Testing step: testing the adjusted convolutional neural network model based on the test set to obtain the recognition accuracy of the folding state; If the recognition accuracy is less than the preset accuracy, then sequentially execute the training step, hyperparameter adjustment step, validation step, and testing step again; If the recognition accuracy is greater than or equal to the preset accuracy, the model training ends to obtain a folding state determination model.

[0066] Specifically, use the validation set data to evaluate the model performance. Input the images marked as "fully folded" into the model, and the output result should be "fully folded". Input the images marked as "partially folded" into the model, and the output result should be "partially folded". Input the images marked as "unfolded" into the model, and the output result should be "unfolded". Monitor the loss and accuracy rate on the training and validation sets, adjust the hyperparameters according to the results, and optimize the model performance. Finally, use the test set data to evaluate the model performance to obtain the recognition accuracy of the final model.

[0067] It should be noted that the hyperparameters include the learning rate, batch size, etc.; the preset accuracy is generally 99%, and it can also be set according to specific requirements.

[0068] In specific practice, step S14, "inputting the real-time acquired image data into the folding state determination model to obtain the folding state of the rearview mirror" includes: inputting the image data collected in real time by the left and right rearview mirror cameras into the folding state determination model to obtain the real-time folding state of the rearview mirror, or, if there is occlusion in the image data collected in real time by the left and right rearview mirror cameras, inputting the real-time acquired bird's-eye view of the vehicle body into the folding state determination model to obtain the real-time folding state of the rearview mirror.

[0069] It should be noted that generally, only the image from the camera on the outer rearview mirror needs to be input to determine whether the rearview mirror is folded. If the image features in this image are occluded, then use an alternative solution. The alternative solution is to input the bird's-eye view generated by the four-way surround cameras and determine whether the camera is folded based on the bird's-eye view.

[0070] In specific practice, the method further includes: if the real-time folding state of the rearview mirror is empty, determining the real-time folding state of the rearview mirror through the angle sensor of the rearview mirror.

[0071] In specific practice, the method further includes: if the real-time folding state of the rearview mirror is empty, determining the real-time folding state of the rearview mirror through the feedback of the folding motor of the rearview mirror.

[0072] Please refer to Figure 2 , Figure 2 which is a schematic block diagram of a device for identifying the folding state of a rearview mirror shown in an exemplary embodiment of the present invention. Refer to Figure 2 , the device 100 for identifying the folding state of the rearview mirror includes:

[0073] An acquisition module 101, configured to acquire and preprocess image data to obtain training set data, validation set data, and test set data; the image data includes image data collected by left and right rearview mirror cameras and a bird's-eye view of the vehicle body in a preset scenario; the preset scenario includes an environmental scenario formed by combining a preset weather, a preset vehicle state, and a preset time point; the preset weather includes sunny, rainy, snowy, and foggy days; the preset vehicle state includes stationary, slow driving, and normal driving; the preset time point is each hour of the 24 hours of a day, and each hour is a preset time point.

[0074] A model training module 102, configured to train, validate, and test a convolutional neural network model based on the training set data, validation set data, and test set data to obtain a folding state determination model.

[0075] A deployment module 103, configured to deploy the folding state determination model to an in-vehicle assistance system.

[0076] A folding state determination module 104, configured to, when the in-vehicle assistance system identifies the current folding state of the rearview mirror, input the real-time acquired image data into the folding state determination model to obtain the folding state of the rearview mirror.

[0077] It should be noted that the technical solutions provided in this embodiment can be applied to scenarios including but not limited to: intelligent driving, automatic parking assistance system, vehicle automatic parking, and parking safety judgment in specific practice.

[0078] Specifically, the image data includes image data collected by left and right rearview mirror cameras and a bird's-eye view of the vehicle body in a preset scenario; the preset scenario includes an environmental scenario formed by combining a preset weather, a preset vehicle state, and a preset time point; the preset weather includes sunny, rainy, snowy, and foggy days; the preset vehicle state includes stationary, slow driving, and normal driving; the preset time point is each hour of the 24 hours of a day, and each hour is a preset time point.

[0079] It can be understood that the device provided in this embodiment obtains and preprocesses image data to obtain training set data, validation set data, and test set data. After training, validating, and testing a convolutional neural network model based on the training set data, validation set data, and test set data, a folding state determination model is obtained. The folding state determination model is deployed in the vehicle-mounted auxiliary system. When the vehicle-mounted auxiliary system recognizes the current folding state of the rearview mirror, the image data obtained in real time is input into the folding state determination model to obtain the folding state of the rearview mirror. This method is based on the image recognition technology of deep learning and combines the image data collected by the left and right rearview mirror cameras and the bird's-eye view of the vehicle body in various environments, and can accurately recognize the folding state of the rearview mirror in various environments.

[0080] Please refer to Figure 3 , Figure 3 which is a schematic block diagram of an electronic device shown in an exemplary embodiment of the present invention. Refer to Figure 3 , the electronic device 200 includes:

[0081] at least one processor 202; and

[0082] a memory 201 communicatively connected to the at least one processor 202; wherein,

[0083] the memory 201 stores instructions executable by the at least one processor 202, and the instructions are executed by the at least one processor 202 to enable the at least one processor 202 to execute any of the above methods for recognizing the folding state of the rearview mirror.

[0084] The above embodiments only represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

[0085] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0086] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0087] The above-described embodiments only represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.

Claims

1. A method for identifying the folding state of a rearview mirror, characterized in that: The method comprises: The image data are acquired and preprocessed to obtain training set data, validation set data and test set data; the image data include image data collected by left and right rearview mirror cameras in a preset scene, and a bird's-eye view of the vehicle body; the preset scene includes an environmental scene formed by combining preset weather, preset vehicle state and preset time point; the preset weather includes sunny, rainy, snowy and foggy days; the preset vehicle state includes stationary, slow driving and normal driving; the preset time point is a preset time point for each hour of 24 hours a day; The convolutional neural network model is trained, verified and tested according to the training set data, the validation set data and the test set data to obtain a folding state determination model; Deploying the folding state determination model into a vehicle-mounted assistance system; When the vehicle-mounted auxiliary system identifies the current folding state of the rearview mirror, the image data acquired in real time is input into the folding state determination model to obtain the folding state of the rearview mirror.

2. The method according to claim 1, characterized in that The image data is obtained and preprocessed to obtain training set data, verification set data and test set data, including: Acquire image data collected by left and right rearview mirror cameras and a bird's-eye view of the vehicle body in a preset scene as the image data; De-noising all images in the image data to obtain target image data; All images in the target image data are annotated one by one and divided into training set data, validation set data and test set data; the training set data accounts for a first preset proportion of the target image data; the validation set data accounts for a second preset proportion of the target image data; and the test set data accounts for a third preset proportion of the target image data.

3. The method according to claim 2, characterized in that The step of performing denoising on all images in the image data to obtain target image data comprises: Performing filtering operations on all images in the image data by using a filtering algorithm so as to reduce noise interference caused by weather and light; Adjust the brightness and contrast of all images in the image data to ensure consistency across all images during model training.

4. The method according to claim 2, characterized in that: The step of labeling all images in the target image data one by one includes: All images in the target image data are annotated one by one by using an annotation tool; the annotated folding states include fully folded, partially folded, and unfolded.

5. The method according to claim 4, characterized in that The folding state determination model is obtained after training, verifying and testing the convolutional neural network model according to the training set data, the validation set data and the test set data, including: Training step: training the convolutional neural network model according to the training set data to obtain a training result; Adjusting hyperparameters step: adjusting the hyperparameters of the convolutional neural network model according to the training results; Verification step: performing model verification on the adjusted convolutional neural network model according to the verification set data; the model verification is verification based on loss and folded state recognition accuracy; Testing step: testing the adjusted convolutional neural network model according to the test set to obtain the recognition accuracy of the folded state; If the recognition accuracy is less than the preset accuracy, the training step, the hyperparameter adjustment step, the verification step and the testing step are performed again in sequence; If the recognition accuracy is greater than or equal to the preset accuracy, the model training is completed to obtain the folding state determination model.

6. The method according to claim 5, characterized in that The step of inputting the image data acquired in real time into the folding state determination model to obtain the folding state of the rearview mirror comprises: The image data collected in real time by the left and right rearview mirror cameras are input into the folding state determination model to obtain the real-time folding state of the rearview mirror, or, If there is occlusion in the image data collected in real time by the left and right rearview mirror cameras, the real-time bird's-eye view of the vehicle body is input into the folding state determination model to obtain the real-time folding state of the rearview mirror.

7. The method according to claim 6, characterized in that The method further comprises: If the real-time folding state of the rearview mirror is empty, the real-time folding state of the rearview mirror is determined by an angle sensor of the rearview mirror.

8. The method according to claim 6, characterized in that The method further comprises: If the real-time folding state of the rearview mirror is empty, the real-time folding state of the rearview mirror is determined through feedback from the folding motor of the rearview mirror.

9. A device for identifying the folding state of a rearview mirror, characterized in that: The device comprises: an acquisition module, for acquiring and preprocessing image data to obtain training set data, validation set data and test set data; the image data includes image data collected by left and right rearview mirror cameras in a preset scene, and a bird's-eye view of the vehicle body; the preset scene includes an environmental scene formed by combining preset weather, preset vehicle state and preset time point; the preset weather includes sunny, rainy, snowy and foggy days; the preset vehicle state includes stationary, slow driving and normal driving; the preset time point is a preset time point for each hour of 24 hours a day; A model training module, used to train, verify and test the convolutional neural network model according to the training set data, the validation set data and the test set data to obtain a folding state determination model; A deployment module, used for deploying the folding state determination model into a vehicle auxiliary system; The folding state determination module is used to obtain the folding state of the rearview mirror after inputting the image data acquired in real time into the folding state determination model when the vehicle-mounted auxiliary system identifies the current folding state of the rearview mirror.

10. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method for identifying the folding state of a rearview mirror as described in any one of claims 1 to 8.

Images