Vehicle parts damage assessment method, device, electronic device and storage medium

Abstract

The vehicle accessory damage assessment method, device, electronic device and storage medium of the present application obtain a set of pictures of a target accessory; input the set of pictures of the target accessory into a pre-trained recognition model to output the recognition result of the target accessory; obtain a judgment value of the target accessory according to the integrity of the target accessory and the recognition accuracy of the recognition model; if the judgment value is less than or equal to the first preset threshold, then according to the judgment value, accessory price, accessory type and accessory material of the target accessory, use a preset accessory residual value algorithm to obtain the residual value result of the target accessory; through the above method, the automatic calculation of the residual value of the target accessory is realized, the accuracy of the residual value calculation of the target accessory is improved, and thus it is beneficial to reduce the compensation cost.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular, to a method, device, electronic device, and storage medium for determining the damage of vehicle parts. Background Art

[0002] In the claim settlement process of vehicle insurance, vehicle damage assessment is an important link. Vehicle damage assessment refers to a verification and confirmation process of the damage situation of on-site accident vehicles. Vehicle damage assessment includes the following links: taking accident photos, finding out the damaged parts of the vehicle, confirming the parts to be replaced or repaired, and giving the final loss amount. Among them, the deduction of residual value is a component of the loss amount. After replacing parts, there are two ways to handle the damaged parts to be replaced, namely, recycling of old parts and selling the residual value. The aforementioned deduction of residual value is the amount of selling the residual value.

[0003] In the prior art, there is a lack of a unified standard for the method of deducting residual value, and most rely on the experience judgment of loss assessors. Due to large subjective judgment differences, there is a risk of leakage, resulting in a low accuracy rate of residual value calculation, and further increasing the claim settlement cost. Summary of the Invention

[0004] The purpose of this application is to provide a method, device, electronic device, and storage medium for determining the damage of vehicle parts to solve the technical problem of low accuracy rate of residual value calculation of vehicle parts in the prior art.

[0005] The technical solution of this application is as follows: A method for determining the damage of vehicle parts is provided, including:

[0006] Obtain a set of pictures of the target part, where the set of pictures includes multiple images to be processed, and the shooting area of at least one of the images to be processed covers the damaged part of the target part;

[0007] Input the set of pictures of the target part into a pre-trained recognition model, and output the recognition result of the target part, where the recognition result includes the part type, part material, and part integrity, and the recognition model is trained according to the sample part images marked with sample types, sample materials, and sample integrity;

[0008] Obtain the judgment value of the target part according to the part integrity of the target part and the recognition accuracy rate of the recognition model;

[0009] If the judgment value is less than or equal to the first preset threshold, then according to the judgment value, part price, part type, and part material of the target part, use the preset part residual value algorithm to obtain the residual value result of the target part.

[0010] In some embodiments, the obtaining a set of pictures of the target part includes:

[0011] Obtain multiple captured images of the target accessory;

[0012] Use a target detection algorithm to perform target detection on the captured images to obtain the detected captured images. The detected captured images include the bounding boxes of the target accessory, and the bounding boxes are the outer regions that frame the target accessory;

[0013] Crop the detected captured images according to the bounding boxes to obtain the corresponding images to be processed, and construct the picture set of the target accessory according to the multiple images to be processed.

[0014] In some embodiments, the images to be processed are divided into at least one category according to the shooting angles of the target accessory;

[0015] Inputting the picture set of the target accessory into a pre-trained recognition model and outputting the recognition result of the target accessory includes:

[0016] Divide the images to be processed into different regions according to a preset division method, respectively extract the first features of different regions in the images to be processed, and output the first image feature matrix of the images to be processed. The first image feature matrix includes the first features of different regions;

[0017] Multiply the first image feature matrix by a preset weight matrix to obtain the second image feature of the images to be processed, where the weight matrix includes the weights of different regions, and the second image feature includes the second features of different regions;

[0018] Obtain the feature matrix of the target accessory, and the feature matrix includes the second features of different regions at different shooting angles of the target accessory;

[0019] Output the recognition result according to the feature matrix of the target accessory.

[0020] In some embodiments, before obtaining the judgment value of the target accessory according to the integrity of the accessory of the target accessory and the recognition accuracy of the recognition model, it further includes:

[0021] Obtain a data set, where the data set includes picture sets of different accessories. The picture set includes multiple images to be processed. The images to be processed are divided into at least one category according to the shooting angles of the accessories, and the shooting regions of at least one of the images to be processed cover the damaged parts of the accessories;

[0022] Divide the dataset into a training set and a test set, and label the images to be processed in the image set of each accessory in the training set for sample type, sample material, and sample integrity.

[0023] Use the training set to train the recognition model to obtain a trained recognition model, where the activation function of the recognition model is f(x) = max(0, w t x + b), x is the first image feature matrix of the image to be processed, w is the weight matrix, and b is a preset parameter.

[0024] Use the test set to test the trained recognition model, and obtain the recognition accuracy of the recognition model according to the recognition result of the trained recognition model.

[0025] In some embodiments, after obtaining the image set of the target accessory, it further includes:

[0026] Perform noise reduction processing on each image to be processed in the image set to obtain a noise-reduced image set.

[0027] Perform data augmentation on the images to be processed in the noise-reduced image set to obtain a data-augmented image set.

[0028] Stitch the images to be processed in the data-augmented image set to obtain a panoramic stitched image of the target accessory.

[0029] Obtain a plurality of standard images to be processed of the target accessory according to the panoramic stitched image, and replace the images to be processed with the standard images to be processed to update the image set of the target accessory, where each standard image to be processed corresponds to a preset shooting angle.

[0030] In some embodiments, the obtaining the salvage value result of the target accessory by using a preset accessory salvage value algorithm according to the judgment value, accessory price, accessory type, and accessory material of the target accessory includes:

[0031] Obtain the corresponding type preset value according to the accessory type.

[0032] Obtain the corresponding material preset value according to the accessory material.

[0033] Obtain the salvage value result of the target accessory by using a preset accessory salvage value algorithm according to the judgment value, accessory price, type preset value corresponding to the accessory type, and material preset value corresponding to the accessory material of the target accessory.

[0034] In some embodiments, after obtaining the judgment value of the target accessory according to the integrity of the accessory and the recognition accuracy of the recognition model, the method further includes:

[0035] If the judgment value is greater than the first preset threshold, a recycling result is output according to the integrity of the target accessory.

[0036] Another technical solution of the present application is as follows: A vehicle accessory damage assessment device is provided, including:

[0037] An acquisition module, configured to acquire a set of pictures of a target accessory, where the set of pictures includes a plurality of images to be processed, and the shooting area of at least one of the images to be processed covers the damaged part of the target accessory;

[0038] A recognition module, configured to input the set of pictures of the target accessory into a pre-trained recognition model, and output a recognition result of the target accessory, where the recognition result includes the accessory type, accessory material, and accessory integrity, and the recognition model is trained according to sample accessory images marked with sample types, sample materials, and sample integrity;

[0039] A first calculation module, configured to obtain a judgment value of the target accessory according to the integrity of the target accessory and the recognition accuracy of the recognition model;

[0040] A second calculation module, configured to, if the judgment value is less than or equal to the first preset threshold, obtain a salvage value result of the target accessory by using a preset accessory salvage value algorithm according to the judgment value, accessory price, accessory type, and accessory material of the target accessory.

[0041] Another technical solution of the present application is as follows: An electronic device is provided, including a processor and a memory coupled to the processor, where the memory stores program instructions executable by the processor; when the processor executes the program instructions stored in the memory, the above vehicle accessory damage assessment method is implemented.

[0042] Another technical solution of the present application is as follows: A storage medium is provided, where program instructions are stored in the storage medium, and when the program instructions are executed by a processor, the above vehicle accessory damage assessment method can be implemented.

[0043] The beneficial effects of the present application are as follows: For the vehicle parts damage assessment method, device, electronic device and storage medium of the present application, a set of pictures of the target part is obtained; the set of pictures of the target part is input into a pre-trained recognition model, and the recognition result of the target part is output; a judgment value of the target part is obtained according to the integrity of the target part and the recognition accuracy of the recognition model; if the judgment value is less than or equal to the first preset threshold, then according to the judgment value, parts price, part type and part material of the target part, the residual value result of the target part is obtained by using a preset parts residual value algorithm; through the above method, automatic calculation of the residual value of the target part is realized, the accuracy of the residual value calculation of the target part is improved, and thus it is beneficial to reduce the compensation cost. Description of the Drawings

[0044] Figure 1 It is a schematic flowchart of the vehicle parts damage assessment method according to an embodiment of the present application;

[0045] Figure 2 It is a schematic structural diagram of the recognition model in an embodiment of the present application;

[0046] Figure 3 It is a schematic principle diagram of the image division method in an embodiment of the present application;

[0047] Figure 4 It is a schematic structural diagram of the vehicle parts damage assessment device according to an embodiment of the present application;

[0048] Figure 5 It is a schematic structural diagram of the electronic device according to an embodiment of the present application;

[0049] Figure 6 It is a schematic structural diagram of the storage medium according to an embodiment of the present application. Detailed Embodiments

[0050] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0051] The terms "first", "second", and "third" in this application are only used for descriptive purposes and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first", "second", and "third" may explicitly or implicitly include at least one of such features. In the description of this application, the meaning of "a plurality" is at least two, such as two, three, etc., unless otherwise specifically defined. In the embodiments of this application, all directional indications (such as up, down, left, right, front, back...) are only used to explain the relative positional relationship and movement conditions between components in a specific posture (as shown in the drawings). If this specific posture changes, then the directional indications will change accordingly. In addition, the terms "comprise" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products, or devices.

[0052] Reference to "embodiment" in this context means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of this application. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0053] An embodiment of this application provides a method for determining the loss of vehicle parts. The execution subject of the method for determining the loss of vehicle parts includes, but is not limited to, at least one of electronic devices such as a server, a terminal, etc. that can be configured to execute the method for determining the loss of vehicle parts provided in the embodiments of this application. In other words, the method for determining the loss of vehicle parts can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to: a single server, a server cluster, a cloud server, or a cloud server cluster, etc.

[0054] Please refer to Figure 1 As shown, it is a flowchart of the method for determining the loss of vehicle parts provided by an embodiment of this application. It should be noted that if there are substantially the same results, the method of this application is not limited to Figure 1 the flowchart order shown. In this embodiment, the method for determining the loss of vehicle parts includes the following steps:

[0055] S10, obtain a set of pictures of the target part, where the set of pictures includes a plurality of images to be processed, and the shooting area of at least one of the images to be processed covers the damaged part of the target part.

[0056] In the embodiments of the present application, the target part is a damaged vehicle part with a damaged area. In some embodiments, the images to be processed in the image set may include at least one of static images and video streams. Exemplarily, the static images contain the target part, or each frame of the video stream contains the target part.

[0057] In some embodiments, step S10 specifically includes the following steps:

[0058] S11, obtaining multiple captured images of the target part;

[0059] Among them, the captured images can be static images or video streams captured by the loss adjuster through a mobile terminal.

[0060] S12, performing object detection on the captured images by using an object detection algorithm to obtain the captured images after detection, and the captured images after detection include a bounding box of the target part, and the bounding box is an external area that frames out the target part;

[0061] In some embodiments, the object detection algorithm is the Mask R-CNN algorithm. By using the Mask R-CNN algorithm to perform object detection on the captured images, a bounding box of the output target part is obtained. That is to say, if the captured image is a static image, in the static image after detection, the target part is extracted by the bounding box, and the image corresponding to the bounding box is the regional image where the target part is located; if the captured image is a video stream, in each frame of the video stream after detection, the target part is extracted by the bounding box, and the image corresponding to the bounding box is the regional image where the target part is located, and the bounding box can extract the image of the target part in each frame of the video stream.

[0062] S13, cropping the captured images after detection according to the bounding box to obtain the corresponding images to be processed, and constructing the image set of the target part according to the multiple images to be processed;

[0063] In some embodiments, the bounding box can be the minimum circumscribed rectangle that frames out the target part to be recognized, so as to remove the influence of the environment as much as possible on the premise of completely obtaining rich detailed features of the target part.

[0064] Since the shooting angles selected by the loss adjuster when holding the mobile terminal to take pictures cannot be unified, and the loss adjuster will select a shooting angle that is conducive to showing the damaged area in order to take a clearer picture of the damaged area, resulting in a large difference in the shooting angles of each target part. To solve the above problems, as an embodiment, after step S10, the following steps are further included:

[0065] S14. Denoise each of the to-be-processed images in the image set to obtain a denoised image set;

[0066] Among them, in order to improve the clarity of the to-be-processed images, a denoising operation is performed on each to-be-processed image in the image set. Specifically, a filter can be used to perform the denoising operation on each to-be-processed image, and the filter can be a median filter or the like.

[0067] S15. Perform data augmentation on the to-be-processed images in the denoised image set to obtain a data-augmented image set;

[0068] Among them, horizontal flipping can be used to perform data augmentation on the to-be-processed images to obtain to-be-processed images with more shooting angles.

[0069] S16. Stitch the to-be-processed images in the data-augmented image set to obtain a panoramic stitched image of the target accessory;

[0070] As an implementation, the panoramic three-dimensional model technology in the prior art can be used for stitching. The panoramic three-dimensional model and stitching information of the target accessory can be obtained in advance, or the panoramic three-dimensional model and stitching information of the target accessory can also be obtained while obtaining the image set. Subsequently, when stitching, the phase algorithm can be used to match each to-be-processed image according to the panoramic three-dimensional model to obtain a panoramic stitched image of the target accessory. The obtained panoramic stitched image is a three-dimensional panoramic image of the target accessory.

[0071] S17. Obtain multiple standard to-be-processed images of the target accessory according to the panoramic stitched image, and replace the to-be-processed images with the standard to-be-processed images to update the image set of the target accessory, where each standard to-be-processed image corresponds to a preset shooting angle.

[0072] Among them, after obtaining the three-dimensional panoramic stitched image, standard to-be-processed images corresponding to the preset shooting angles can be intercepted from the panoramic stitched image according to the preset shooting angles, and the original to-be-processed images in the image set are replaced with the standard to-be-processed images. In the subsequent recognition step, the target accessory is recognized according to the standard to-be-processed images.

[0073] S20. Input the image set of the target accessory into a pre-trained recognition model, and output the recognition result of the target accessory, where the recognition result includes the accessory type, accessory material, and accessory integrity, and the recognition model is trained according to the sample accessory images labeled with the sample type, sample material, and sample integrity.

[0074] In some embodiments, the recognition model is a Convolutional Neural Network (CNN). The recognition model may include several convolutional layers and several fully connected layers. Among them, a convolutional layer (Conv) refers to a layered structure composed of several convolutional units in a convolutional neural network layer. A convolutional neural network is a feedforward neural network, which includes at least two neural network layers. Each neural network layer contains several neurons, and the neurons are arranged in layers. Neurons in the same layer are not connected to each other, and the transmission of information between layers only proceeds in one direction. A fully connected layer (FC) means that each node in this layered structure is connected to all nodes in the previous layer, and can be used to comprehensively process the features extracted by the neural network layer in the previous layer, playing the role of a "classifier" in the neural network model.

[0075] In some embodiments, the recognition model may further include a batch normalization layer, an activation function layer, and a pooling layer. Among them, a batch normalization layer (BN) refers to a layered structure that can unify scattered data, making the data input into the neural network model have a unified specification, making it easier for the neural network model to find patterns in the data, and can optimize the neural network model. An activation function layer (AF) refers to a layered structure of functions operating on neurons in the neural network model, which can map the input of neurons to the output end. By introducing a non-linear function into the neural network model, the output value of the neural network model can approximate any non-linear function. The pooling layer is also named the sampling layer. After the convolutional layer, it refers to a layered structure that can extract features from the input values again. The pooling layer can ensure the main features of the values in the previous layer, and can also reduce the parameters and computational amount in the next layer. The pooling layer consists of multiple feature maps. One feature map in the convolutional layer corresponds to one feature map in the pooling layer, and the number of feature maps will not change. Features with spatial invariance are obtained by reducing the resolution of the feature maps.

[0076] Specifically, Figure 2 The structural schematic diagram of the recognition model provided for an exemplary embodiment of this application is shown in Figure 2As shown, the recognition model includes an input layer, two convolutional modules, and three fully connected modules. Each convolutional module includes at least one convolutional layer. Further, each convolutional module may also include a batch normalization layer, an activation function layer, or a pooling layer. Each fully connected module includes at least one fully connected layer, and the three fully connected modules respectively correspond to the accessory type, accessory material, and accessory integrity. Among them, the convolutional layer of the first convolutional module is used to extract features from each image to be processed in the image set of the target accessory. The convolutional layer of the second convolutional module is used to perform a convolutional operation on the output of the first convolutional module. The first fully connected module is used to obtain the probability value of the target accessory being each preset accessory type according to the output of the second convolutional module. For example, the preset accessory types include headlight, bumper, turn signal, front cover, rear cover, side trim panel, panel assembly, etc. The second fully connected module is used to obtain the probability value of the target accessory being each preset accessory material according to the output of the second convolutional module. For example, the preset accessory materials include copper, iron, aluminum, plastic, rubber, etc. The third fully connected module is used to obtain the integrity of the target accessory according to the output of the second convolutional module. For example, the integrity of the target accessory can be 90%, 89%, 80%, etc. Among them, the integrity of a brand-new accessory is 100%. Due to deformation or missing of the damaged part, the integrity of the target accessory is less than 100%.

[0077] As an implementation manner, the images to be processed are divided into at least one category according to the shooting angle of the target accessory. The shooting angle may include, but is not limited to, a front view angle, a left view angle, a right view angle, a top view angle, or a bottom view angle, etc. Correspondingly, step S20 specifically includes the following steps:

[0078] S21, divide the images to be processed into different regions according to a preset division method, respectively extract the first features of different regions in the images to be processed, and output the first image feature matrix of the images to be processed. The first image feature matrix includes the first features of different regions;

[0079] As an implementation manner, the preset division method may be Figure 3 As shown, divide the image to be processed into image blocks of the same size. Each image block corresponds to a region. When extracting features from the image to be processed, extract features for each image block respectively to obtain the first feature of each image block. The first image feature matrix F i =[F i 1 , F i 2 , …, F i j , …, F i N , N is the number of image blocks, and F ij is the first feature of the corresponding area, 1 ≤ j ≤ N, the i-th image to be processed corresponds to the i-th shooting angle, 1 ≤ i ≤ M, and M is the number of shooting angles.

[0080] S22, multiply the first image feature matrix by a preset weight matrix to obtain the second image feature of the image to be processed, where the weight matrix includes the weights of different regions, and the second image feature includes the second features of different regions;

[0081] Among them, each weight element in the weight matrix corresponds to each image block, and the weight matrix W i = [w i 1 , w i 2 , …, w i j , …, w i N , N is the number of image blocks, and w i j is the weight of the corresponding area, 1 ≤ j ≤ N. The first image feature matrix B i = [B i 1 , B i 2 , …, B i j , …, B i N , N is the number of image blocks, and B i j is the second feature of the corresponding area, B i j = F i j * w i j , 1 ≤ j ≤ N.

[0082] S23, obtain the feature matrix of the target accessory, where the feature matrix includes the second features of different regions of different shooting angles of the target accessory;

[0083] Among them, the second image features of different images to be processed of the target accessory are merged to obtain the feature matrix T of the target accessory = [B1 1 , …, B1 N , …, B i 1 , …, B i N , …, B M 1 , …, B M N .

[0084] S24. Output the recognition result according to the feature matrix of the target accessory.

[0085] In some embodiments, continue to perform a convolution operation and a fully connected operation on the feature matrix of the target accessory in sequence to output a recognition result.

[0086] S30. Obtain a judgment value of the target accessory according to the integrity of the target accessory and the recognition accuracy of the recognition model.

[0087] As an embodiment, the judgment value of the target accessory can be calculated in the following way: Judgment value X = Integrity a / Accuracy b. For example, if the integrity a output by the recognition model is 80% and the recognition accuracy of the recognition model is 90%, the corresponding judgment value is 80% / 90% = 89%.

[0088] In this embodiment, the judgment value is used to indicate whether the target accessory is to be recycled or have its damaged parts deducted.

[0089] As another embodiment, before step S20, the following steps are further included:

[0090] S31. Obtain a data set, where the data set includes a set of pictures of different accessories. The set of pictures includes multiple images to be processed. The images to be processed are divided into at least one category according to the shooting angle of the accessories, and the shooting area of at least one of the images to be processed covers the damaged part of the accessory;

[0091] S32. Divide the data set into a training set and a test set, and label the images to be processed in the set of pictures of each accessory in the training set with sample types, sample materials, and sample integrity;

[0092] S33. Use the training set to train the recognition model to obtain a trained recognition model, where the activation function of the recognition model is f(x) = max(0, w t x + b), x is the first image feature matrix of the image to be processed, w is the weight matrix, and b is a preset parameter;

[0093] S34. Use the test set to test the trained recognition model, and obtain the recognition accuracy of the recognition model according to the recognition result of the trained recognition model.

[0094] As a preferred embodiment, in order to more accurately label the integrity, in step S31, when obtaining the data set, if the shooting area covers the damaged part, when using the electronic device to obtain the shooting image corresponding to the shooting area, the shooting area of the accessory is scanned with infrared rays at the same time to obtain the surface structure information of the damaged part of the accessory, and then the damage degree of the damaged part is obtained according to the shooting image corresponding to the shooting area and the surface structure information of the damaged part, and further the integrity of the accessory is obtained for labeling.

[0095] S40. If the judgment value is less than or equal to the first preset threshold, then according to the judgment value, accessory price, accessory type and accessory material of the target accessory, the residual value result of the target accessory is obtained by using a preset accessory residual value algorithm.

[0096] Among them, when the judgment value is less than the first preset threshold, it means that the integrity of the target accessory is relatively low and the recycling cost is relatively high, and it is suitable for residual value deduction processing. The residual value result of the target accessory is obtained according to the judgment value, the accessory price of the corresponding target accessory, the accessory type output by the recognition model and the accessory material output by the recognition model.

[0097] As an embodiment, step S40 specifically includes the following steps:

[0098] S41. Obtain the corresponding type preset value according to the accessory type;

[0099] In this embodiment, the type preset value is a correction parameter, and different accessory types correspond to different type preset values. A type parameter table for recording the corresponding relationship between the accessory type and the type preset value can be set in advance, and the corresponding type preset value is queried in the type parameter table according to the accessory type.

[0100] S42. Obtain the corresponding material preset value according to the accessory material;

[0101] In this embodiment, the material preset value is also a correction parameter, and different accessory materials correspond to different material preset values. A material parameter table for recording the corresponding relationship between the accessory material and the material preset value can be set in advance, and the corresponding material preset value is queried in the material parameter table according to the accessory material.

[0102] S43. According to the judgment value of the target accessory, the accessory price, the type preset value corresponding to the accessory type and the material preset value corresponding to the accessory material, the residual value result of the target accessory is obtained by using a preset accessory residual value algorithm.

[0103] As an implementation manner, the preset accessory residual value algorithm is as follows: the residual value result r = judgment value X * type preset value c * accessory price e + material preset value d. That is to say, the residual value result of the target accessory is the sum of the product of the judgment value, the type preset value, and the accessory price and the material preset value.

[0104] In some implementation manners, after step S30, the following steps are further included:

[0105] S50, if the judgment value is greater than the first preset threshold, then output a recycling result according to the integrity of the target accessory.

[0106] As Figure 4 shown, an embodiment of the present application provides a vehicle accessory damage assessment device. The device 40 includes an acquisition module 41, an identification module 42, a first calculation module 43, and a second calculation module 44. Among them, the acquisition module 41 is used to acquire a picture set of the target accessory, where the picture set includes a plurality of images to be processed, and the shooting area of at least one of the images to be processed covers the damaged part of the target accessory; the identification module 42 is used to input the picture set of the target accessory into a pre-trained identification model and output the identification result of the target accessory, where the identification result includes the accessory type, the accessory material, and the accessory integrity, and the identification model is trained according to the sample accessory images marked with sample types, sample materials, and sample integrity; the first calculation module 43 is used to obtain the judgment value of the target accessory according to the integrity of the target accessory and the identification accuracy of the identification model; the second calculation module 44 is used to, if the judgment value is less than or equal to the first preset threshold, use the preset accessory residual value algorithm to obtain the residual value result of the target accessory according to the judgment value, the accessory price, the accessory type, and the accessory material of the target accessory.

[0107] In some implementation manners, the acquisition module 41 is further used to acquire a plurality of shooting images of the target accessory; perform target detection on the shooting images by using a target detection algorithm to obtain the detected shooting images, where the detected shooting images include the bounding box of the target accessory, and the bounding box is the external area that frames the target accessory; crop the detected shooting images according to the bounding box to obtain the corresponding images to be processed, and construct the picture set of the target accessory according to the plurality of images to be processed.

[0108] In some embodiments, the image to be processed is divided into at least one category according to the shooting angle of the target accessory; correspondingly, the recognition module 42 is further configured to divide the image to be processed into different regions according to a preset division method, respectively extract the first features of different regions in the image to be processed, output the first image feature matrix of the image to be processed, where the first image feature matrix includes the first features of different regions; multiply the first image feature matrix by a preset weight matrix to obtain the second image feature of the image to be processed, where the weight matrix includes the weights of different regions, and the second image feature includes the second features of different regions; obtain the feature matrix of the target accessory, where the feature matrix includes the second features of different regions of different shooting angles of the target accessory; output the recognition result according to the feature matrix of the target accessory.

[0109] In some embodiments, the recognition module 42 is further configured to obtain a data set, where the data set includes a set of pictures of different accessories, the set of pictures includes a plurality of images to be processed, the images to be processed are divided into at least one category according to the shooting angle of the accessories, and the shooting area of at least one of the images to be processed covers the damaged part of the accessories; divide the data set into a training set and a test set, and label the images to be processed in the set of pictures of each accessory in the training set with sample types, sample materials, and sample integrity; use the training set to train the recognition model to obtain a trained recognition model, where the activation function of the recognition model is f(x) = max(0, w t x + b), x is the first image feature matrix of the image to be processed, w is the weight matrix, and b is a preset parameter; use the test set to test the trained recognition model, and obtain the recognition accuracy of the recognition model according to the recognition result of the trained recognition model.

[0110] In some embodiments, the acquisition module 41 is further configured to perform noise reduction processing on each image to be processed in the set of pictures to obtain a set of pictures after noise reduction processing; perform data augmentation on the images to be processed in the set of pictures after noise reduction processing to obtain a set of pictures after data augmentation; splice the images to be processed in the set of pictures after data augmentation to obtain a panoramic mosaic image of the target accessory; obtain a plurality of standard images to be processed of the target accessory according to the panoramic mosaic image, and replace the images to be processed with the standard images to be processed to update the set of pictures of the target accessory, where each standard image to be processed corresponds to a preset shooting angle.

[0111] In some embodiments, the second calculation module 44 is further configured to obtain a corresponding type preset value according to the accessory type; obtain a corresponding material preset value according to the accessory material; and obtain a salvage result of the target accessory by using a preset accessory salvage algorithm based on the judgment value, the accessory price, the type preset value corresponding to the accessory type, and the material preset value corresponding to the accessory material.

[0112] In some embodiments, the second calculation module 44 is further configured to, if the judgment value is greater than the first preset threshold, output a recycling result according to the integrity of the target accessory.

[0113] Figure 5 is a schematic structural diagram of an electronic device according to an embodiment of the present application. As Figure 5 shown, the electronic device 50 includes a processor 51 and a memory 52 coupled to the processor 51.

[0114] The memory 52 stores program instructions for implementing the vehicle accessory damage assessment method according to any one of the above embodiments.

[0115] The processor 51 is configured to execute the program instructions stored in the memory 52 to perform vehicle accessory damage assessment.

[0116] Among them, the processor 51 may also be referred to as a CPU (Central Processing Unit, central processing unit). The processor 51 may be an integrated circuit chip with signal processing capabilities. The processor 51 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0117] Refer to Figure 6 , Figure 6The figure shows a schematic structure of a storage medium according to an embodiment of the present application. The storage medium 60 in the embodiment of the present application stores program instructions 61 that can implement all of the above methods. Among them, the program instructions 61 can be stored in the above storage medium in the form of a software product, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods described in various embodiments of the present application. The aforementioned storage medium includes: various media that can store program codes such as USB flash drives, external hard drives, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs, or terminal devices such as computers, servers, mobile phones, and tablets.

[0118] In several embodiments provided by the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of devices or units can be in electrical, mechanical, or other forms.

[0119] In addition, in each embodiment of the present application, the functional units can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units. The above is only the embodiment of the present application, and it does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present application, or directly or indirectly applied to other related technical fields, is equally included in the patent protection scope of the present application.

[0120] The above are only the embodiments of the present application. It should be noted here that for those of ordinary skill in the art, improvements can be made without departing from the creative concept of the present application, but these all belong to the protection scope of the present application.

Claims

1. A method for determining the damage of vehicle parts, characterized in that, Including: Obtain a set of pictures of the target accessory, where the set of pictures includes multiple images to be processed, and the shooting area of at least one of the images to be processed covers the damaged part of the target accessory; Input the set of pictures of the target accessory into a pre-trained recognition model, and output the recognition result of the target accessory, where the recognition result includes the accessory type, accessory material, and accessory integrity, and the recognition model is trained according to the sample accessory images marked with sample type, sample material, and sample integrity; Obtain the judgment value of the target accessory according to the accessory integrity of the target accessory and the recognition accuracy of the recognition model; If the judgment value is less than or equal to the first preset threshold, then according to the judgment value, accessory price, accessory type, and accessory material of the target accessory, use the preset accessory salvage value algorithm to obtain the salvage value result of the target accessory; The images to be processed are divided into at least one category according to the shooting angle of the target accessory; The step of inputting the set of pictures of the target accessory into a pre-trained recognition model and outputting the recognition result of the target accessory includes: Divide the images to be processed into different regions according to a preset division method, respectively extract the first features of different regions in the images to be processed, and output the first image feature matrix of the images to be processed, where the first image feature matrix includes the first features of different regions; Multiply the first image feature matrix by a preset weight matrix to obtain the second image feature of the image to be processed, where the weight matrix includes the weights of different regions, and the second image feature includes the second features of different regions; Obtain the feature matrix of the target accessory, where the feature matrix includes the second features of different regions at different shooting angles of the target accessory; Output the recognition result according to the feature matrix of the target accessory.

2. The vehicle accessory damage assessment method according to claim 1, wherein The step of obtaining the set of pictures of the target accessory includes: Obtain multiple shooting images of the target accessory; Use a target detection algorithm to perform target detection on the shooting images to obtain the detected shooting images, where the detected shooting images include the bounding box of the target accessory, and the bounding box is the external area that frames the target accessory; Crop the detected shooting images according to the bounding box to obtain the corresponding images to be processed, and construct the set of pictures of the target accessory according to the multiple images to be processed.

3. The vehicle accessory damage assessment method according to claim 1, wherein Before obtaining the judgment value of the target accessory according to the accessory integrity of the target accessory and the recognition accuracy of the recognition model, it further includes: Obtain a data set, where the data set includes sets of pictures of different accessories, the set of pictures includes multiple images to be processed, the images to be processed are divided into at least one category according to the shooting angle of the accessory, and the shooting area of at least one of the images to be processed covers the damaged part of the accessory; Divide the data set into a training set and a test set, and label the sample type, sample material, and sample integrity of the images to be processed in the set of pictures of each accessory in the training set; Train the recognition model using the training set to obtain a trained recognition model, where the activation function of the recognition model is f(x) = max(0, w t x + b), x is the first image feature matrix of the image to be processed, w is the weight matrix, and b is a preset parameter; Use the test set to test the trained recognition model, and obtain the recognition accuracy rate of the recognition model according to the recognition result of the trained recognition model.

4. The vehicle accessory damage assessment method according to claim 1, characterized in that After obtaining the picture set of the target accessory, it further includes: Perform noise reduction processing on each of the to-be-processed images in the picture set to obtain a picture set after noise reduction processing; Perform data augmentation on the to-be-processed images in the picture set after noise reduction processing to obtain a picture set after data augmentation; Stitch the to-be-processed images in the picture set after data augmentation to obtain a panoramic stitched picture of the target accessory; Obtain a plurality of standard to-be-processed images of the target accessory according to the panoramic stitched picture, and replace the to-be-processed images with the standard to-be-processed images to update the picture set of the target accessory, wherein each standard to-be-processed image corresponds to a preset shooting angle.

5. The vehicle accessory damage assessment method according to claim 1, characterized in that, The obtaining the residual value result of the target accessory by using a preset accessory residual value algorithm according to the judgment value, accessory price, accessory type and accessory material of the target accessory includes: Obtain the corresponding type preset value according to the accessory type; Obtain the corresponding material preset value according to the accessory material; Obtain the residual value result of the target accessory by using a preset accessory residual value algorithm according to the judgment value of the target accessory, the accessory price, the type preset value corresponding to the accessory type and the material preset value corresponding to the accessory material.

6. The vehicle accessory damage assessment method according to claim 1, wherein After obtaining the judgment value of the target accessory according to the accessory integrity of the target accessory and the recognition accuracy rate of the recognition model, it further includes: If the judgment value is greater than the first preset threshold, output a recycling result according to the integrity of the target accessory.

7. A vehicle accessory damage assessment device, characterized in that, It includes: An acquisition module, configured to acquire a picture set of a target accessory, wherein the picture set includes a plurality of to-be-processed images, and the shooting area of at least one of the to-be-processed images covers the damaged part of the target accessory; A recognition module, configured to input the picture set of the target accessory into a pre-trained recognition model, and output a recognition result of the target accessory, wherein the recognition result includes an accessory type, an accessory material, and an accessory integrity, and the recognition model is trained according to sample accessory images labeled with sample types, sample materials, and sample integrity; A first calculation module, configured to obtain a judgment value of the target accessory according to the accessory integrity of the target accessory and the recognition accuracy rate of the recognition model; A second calculation module, configured to, if the judgment value is less than or equal to a first preset threshold, obtain the residual value result of the target accessory by using a preset accessory residual value algorithm according to the judgment value, accessory price, accessory type and accessory material of the target accessory; The to-be-processed images are divided into at least one category according to the shooting angle of the target accessory; The recognition module is further configured to: Divide the to-be-processed image into different regions according to a preset division method, respectively extract the first features of different regions in the to-be-processed image, and output the first image feature matrix of the to-be-processed image, where the first image feature matrix includes the first features of different regions; Multiply the first image feature matrix by a preset weight matrix to obtain the second image feature of the to-be-processed image, where the weight matrix includes the weights of different regions, and the second image feature includes the second features of different regions; Obtain the feature matrix of the target accessory, where the feature matrix includes the second features of different regions at different shooting angles of the target accessory; Output the recognition result according to the feature matrix of the target accessory.

8. An electronic device, characterized in that, It includes a processor and a memory coupled to the processor, and the memory stores program instructions executable by the processor; when the processor executes the program instructions stored in the memory, it implements the vehicle accessory damage assessment method according to any one of claims 1 to 6.

9. A storage medium, characterized in that, Program instructions are stored in the storage medium, and when the program instructions are executed by a processor, they can implement the vehicle accessory damage assessment method according to any one of claims 1 to 6.

Images