Vehicle damage assessment method and device

Abstract

The embodiments of the present application disclose a vehicle damage assessment method and device. A specific implementation of the method includes: acquiring a vehicle image of a target vehicle, wherein the vehicle image includes a three-dimensional scanned image of the appearance; performing image recognition on the vehicle image to obtain damage information of the target vehicle, wherein the damage information is used to characterize the loss type and degree of damage of the damaged parts; based on the damage information, generating repair method information of the damaged parts of the target vehicle, and generating a damage assessment result including the damage information and the repair method information. The embodiments of the present application can utilize a vehicle image including a three-dimensional scanned image of the appearance, and accurately and efficiently determine damage information including appearance damage through image recognition, thereby improving the accuracy and efficiency of damage assessment.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of computer technology, specifically to the field of Internet technology, and especially to a vehicle damage assessment method and device. Background Art

[0002] In recent years, with the continuous increase in the number of cars on the road, traffic accidents have also increased. Subsequently, the workload of vehicle inspection and damage assessment has also increased.

[0003] In the prior art, manual damage assessment is generally used. During the manual damage assessment process, the workload of personnel is very large and the damage assessment cycle is long. Therefore, this method is not only inefficient and consumes a lot of manpower, but also the damage assessment results obtained are unstable. Summary of the invention

[0004] The embodiments of the present application provide a vehicle damage assessment method and device.

[0005] In a first aspect, an embodiment of the present application provides a vehicle damage assessment method, comprising: acquiring a vehicle image of a target vehicle, wherein the vehicle image includes a three-dimensional scan image of the appearance; performing image recognition on the vehicle image to obtain damage information of the target vehicle, wherein the damage information is used to characterize the loss type and degree of damage of damaged parts; based on the damage information, generating repair method information of the damaged parts of the target vehicle, and generating a damage assessment result including the damage information and the repair method information.

[0006] In some embodiments, the vehicle image includes an appearance shot image, an appearance three-dimensional scanning image and a radiographic detection image; and obtaining a vehicle image of a target vehicle includes: obtaining a distant image of the target vehicle, performing target detection on the distant image, and obtaining damaged part information; for the damaged part indicated by the damaged part information, obtaining a close-up image and an appearance three-dimensional scanning image, and using the close-up image and the distant image as an appearance shot image; obtaining a radiographic detection image of the target vehicle; and using the appearance shot image, the appearance three-dimensional scanning image and the radiographic detection image as the vehicle image.

[0007] In some embodiments, image recognition is performed on a vehicle image to obtain damage information of the target vehicle, including: using three pre-trained semantic segmentation models to respectively process an appearance photographed image, an appearance 3D scanning image, and a X-ray detection image to obtain sub-damage information output by each semantic segmentation model, wherein the sub-damage information includes at least one of the following: damaged accessory information, damage type information of the damaged accessory indicated by the damaged accessory information, and damage degree information of the damaged accessory; and obtaining damage information based on a combination of the sub-damage information respectively obtained by the three semantic segmentation models.

[0008] In some embodiments, three pre-trained semantic segmentation models are used to process appearance captured images, appearance three-dimensional scanning images and X-ray detection images respectively, including: mapping the near-field image to the distant field image to obtain a mapping result; inputting the shooting angle, damage part information and mapping result of the distant field image into the first semantic segmentation model, wherein the sub-damage information output by the first semantic segmentation model is appearance damage quantification information; inputting the shooting angle, damage part information and appearance three-dimensional scanning image of the distant field image into the second semantic segmentation model, wherein the sub-damage information output by the second semantic segmentation model is appearance three-dimensional damage information; inputting the shooting angle, damage part information and X-ray detection image of the distant field image into the third semantic segmentation model, wherein the sub-damage information output by the third semantic segmentation model is internal damage information.

[0009] In some embodiments, the shooting angle is obtained by the following steps: using a pre-trained angle classification model to classify the shooting angle of the distant view image to obtain the shooting angle of the distant view image.

[0010] In some embodiments, the vehicle image includes an appearance shot image and an appearance three-dimensional scanning image; and obtaining a vehicle image of a target vehicle includes: obtaining a distant image of the target vehicle, performing target detection on the distant image, and obtaining damaged part information; for the damaged part indicated by the damaged part information, obtaining a close-up image and an appearance three-dimensional scanning image, and using the close-up image and the distant image as the appearance shot image; using the appearance shot image and the appearance three-dimensional scanning image as the vehicle image.

[0011] In some embodiments, target detection is performed on a distant image to obtain damaged part information, including: in response to the distant image containing a damaged part, target detection is performed on the distant image to obtain damaged part information indicating the damaged part contained in the distant image, and position information of the damaged part in the distant image; and for the damaged part indicated by the damaged part information, a close-up image and an appearance three-dimensional scanning image are obtained, including: based on the position information, a close-up mode switching instruction is generated, and the close-up mode switching instruction is output to a shooting device; and a close-up image of the damaged part indicated by the damaged part information is obtained.

[0012] In some embodiments, the damage information is text information; and the damage information is obtained by combining the sub-damage information obtained respectively based on the three semantic segmentation models, including: combining the sub-damage information obtained respectively by the three semantic segmentation models into damage information to be processed; using a pre-trained text output model to convert the damage information to be processed into damage information.

[0013] In some embodiments, based on the damage information, repair method information of the damaged parts of the target vehicle is generated, including: inputting the basic vehicle information and damage information of the target vehicle into a pre-trained repair method classification model, and obtaining the repair method of the damaged parts indicated by the damaged parts information output from the repair method classification model.

[0014] In some embodiments, a damage assessment result including damage information and repair method information is generated, including: querying repair value reference information corresponding to the damage information; determining the repair discount rate of the target vehicle based on the vehicle additional information, and determining the repair value information corresponding to the repair value reference information based on the repair discount rate; inputting the repair value information, repair method information and vehicle basic information into a pre-generated sequence-to-sequence model to obtain text information output from the sequence-to-sequence model; and using the output text information as a damage assessment report for the target vehicle.

[0015] In some embodiments, the method also includes: acquiring an accident scene image of the target vehicle, determining the cause of the accident corresponding to the accident scene image; and inputting the repair value information, repair method information and vehicle basic information into a pre-generated sequence-to-sequence model to obtain text information output from the sequence-to-sequence model, including: inputting the cause of the accident, repair value information, repair method information and vehicle basic information into a pre-generated sequence-to-sequence model to obtain text information output from the sequence-to-sequence model.

[0016] In some embodiments, basic information of each vehicle of the target vehicle is stored in a basic information relationship map.

[0017] In a second aspect, an embodiment of the present application provides a vehicle damage assessment device, comprising: an acquisition unit, configured to acquire a vehicle image of a target vehicle, wherein the vehicle image includes a three-dimensional scanned image of the appearance; an identification unit, configured to perform image recognition on the vehicle image to obtain damage information of the target vehicle, wherein the damage information is used to characterize the loss type and degree of damage of the damaged parts; a generation unit, configured to generate repair method information of the damaged parts of the target vehicle based on the damage information, and generate a damage assessment result including the damage information and the repair method information.

[0018] In some embodiments, the vehicle image includes an exterior shot image, an exterior three-dimensional scanning image and a radiographic detection image; and an acquisition unit, including: an acquisition module, configured to acquire a distant image of the target vehicle, perform target detection on the distant image, and obtain damaged part information; a first image module, configured to acquire a close-up image and an exterior three-dimensional scanning image for the damaged part indicated by the damaged part information, and use the close-up image and the distant image as an exterior shot image; a second image module, configured to acquire a radiographic detection image of the target vehicle; and a third image module, configured to use the exterior shot image, the exterior three-dimensional scanning image and the radiographic detection image as the vehicle image.

[0019] In some embodiments, the recognition unit includes: a separate processing module, configured to use three pre-trained semantic segmentation models to respectively process the appearance shot image, the appearance three-dimensional scanning image and the X-ray detection image, and obtain sub-damage information output by each semantic segmentation model, wherein the sub-damage information includes at least one of the following: damaged accessory information, damage type information of the damaged accessory indicated by the damaged accessory information, and damage degree information of the damaged accessory; a combination module, configured to obtain damage information based on the combination of the sub-damage information obtained by the three semantic segmentation models.

[0020] In some embodiments, the separate processing modules are further configured to: map the near-field image to the distant-field image to obtain a mapping result; input the shooting angle, damage part information and mapping result of the distant-field image into a first semantic segmentation model, wherein the sub-damage information output by the first semantic segmentation model is appearance damage quantification information; input the shooting angle, damage part information and appearance three-dimensional scanning image of the distant-field image into a second semantic segmentation model, wherein the sub-damage information output by the second semantic segmentation model is appearance three-dimensional damage information; input the shooting angle, damage part information and X-ray detection image of the distant-field image into a third semantic segmentation model, wherein the sub-damage information output by the third semantic segmentation model is internal damage information.

[0021] In some embodiments, the vehicle image includes an exterior shot image and an exterior three-dimensional scanning image; and obtaining a vehicle image of a target vehicle includes: a detection module, configured to obtain a distant image of the target vehicle, perform target detection on the distant image, and obtain damaged part information; a fourth image module, configured to obtain a close-up image and an exterior three-dimensional scanning image for the damaged part indicated by the damaged part information, and use the close-up image and the distant image as the exterior shot image; and a fifth image module, configured to use the exterior shot image and the exterior three-dimensional scanning image as the vehicle image.

[0022] In some embodiments, the acquisition module or the detection module is further configured to: in response to the distant image containing a damaged part, perform target detection on the distant image to obtain damaged part information indicating the damaged part contained in the distant image, and position information of the damaged part in the distant image; and the first image module is further configured to: generate a close-up mode switching instruction based on the position information, and output the close-up mode switching instruction to the shooting device; and obtain a close-up image of the damaged part indicated by the damaged part information.

[0023] In some embodiments, the damage information is text information; and the combination module is further configured to: combine the sub-damage information obtained by the three semantic segmentation models into the damage information to be processed; and use a pre-trained text output model to convert the damage information to be processed into damage information.

[0024] In some embodiments, the generation unit includes: a method determination module, which is configured to input the basic vehicle information and damage information of the target vehicle into a pre-trained repair method classification model, and obtain the repair method of the damaged parts indicated by the damaged parts information output from the repair method classification model.

[0025] In some embodiments, the generation unit includes: a query module, configured to query the repair value reference information corresponding to the damage information; a value determination module, configured to determine the repair discount rate of the target vehicle based on the vehicle additional information, and determine the repair value information corresponding to the repair value reference information based on the repair discount rate; a text output module, configured to input the repair value information, repair method information and vehicle basic information into a pre-generated sequence-to-sequence model to obtain text information output from the sequence-to-sequence model; a report generation module, configured to use the output text information as the damage assessment report of the target vehicle.

[0026] In some embodiments, the device also includes: a cause determination unit, configured to obtain an accident scene image of the target vehicle and determine the cause of the accident corresponding to the accident scene image; and a text output module, further configured to: input the cause of the accident, repair value information, repair method information and vehicle basic information into a pre-generated sequence-to-sequence model to obtain text information output from the sequence-to-sequence model.

[0027] In some embodiments, the shooting angle is obtained by the following steps: using a pre-trained angle classification model to classify the shooting angle of the distant view image to obtain the shooting angle of the distant view image.

[0028] In some embodiments, basic information of each vehicle of the target vehicle is stored in a basic information relationship map.

[0029] In a third aspect, an embodiment of the present application provides an electronic device, comprising: one or more processors; a storage device for storing one or more programs, and when the one or more programs are executed by one or more processors, the one or more processors implement a method such as any embodiment of the vehicle damage assessment method.

[0030] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method such as any one of the embodiments of the vehicle damage assessment method.

[0031] The vehicle damage assessment solution provided in the embodiment of the present application first obtains a vehicle image of the target vehicle, wherein the vehicle image includes a three-dimensional scan image of the appearance. Thereafter, image recognition is performed on the vehicle image to obtain damage information of the target vehicle, wherein the damage information is used to characterize the loss type and degree of damage of the damaged parts. Finally, based on the damage information, repair method information of the damaged parts of the target vehicle is generated, and a damage assessment result including the damage information and the repair method information is generated. The embodiment of the present application can utilize a vehicle image including a three-dimensional scan image of the appearance, and accurately and efficiently determine damage information including appearance damage through image recognition, thereby improving the accuracy and efficiency of damage assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0033] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0034] Figure 2 is a flow chart of an embodiment of a vehicle damage assessment method according to the present application;

[0035] Figure 3 is a schematic diagram of an application scenario of the vehicle damage assessment method according to the present application;

[0036] Figure 4 is a flow chart of another embodiment of the vehicle damage assessment method according to the present application;

[0037] Figure 5 is a schematic structural diagram of an embodiment of a vehicle damage assessment device according to the present application;

[0038] Figure 6 It is a structural diagram of a computer system suitable for implementing an electronic device of an embodiment of the present application. DETAILED DESCRIPTION

[0039] The present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It should also be noted that, for ease of description, only the parts related to the relevant invention are shown in the accompanying drawings.

[0040] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0041] Figure 1 An exemplary system architecture 100 is shown to which an embodiment of a vehicle damage assessment method or a vehicle damage assessment device of the present application can be applied.

[0042] like Figure 1 As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0043] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as image processing applications, video applications, live broadcast applications, instant messaging tools, email clients, social platform software, etc.

[0044] The terminal devices 101, 102, 103 here can be hardware or software. When the terminal devices 101, 102, 103 are hardware, they can be various electronic devices with display screens, including but not limited to smart phones, tablet computers, e-book readers, laptop portable computers and desktop computers, etc. When the terminal devices 101, 102, 103 are software, they can be installed in the electronic devices listed above. It can be implemented as multiple software or software modules (for example, multiple software or software modules for providing distributed services), or it can be implemented as a single software or software module. No specific limitation is made here.

[0045] The server 105 may be a server that provides various services, such as a background server that provides support for the terminal devices 101, 102, and 103. The background server may analyze and process the received data such as vehicle images, and feed back the processing results (such as damage assessment results) to the terminal device.

[0046] It should be noted that the vehicle damage assessment method provided in the embodiment of the present application can be executed by the server 105 or the terminal devices 101, 102, 103, and accordingly, the vehicle damage assessment device can be set in the server 105 or the terminal devices 101, 102, 103.

[0047] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0048] Continue to refer Figure 2 , shows a process 200 of an embodiment of a vehicle damage assessment method according to the present application. The vehicle damage assessment method comprises the following steps:

[0049] Step 201, obtaining a vehicle image of a target vehicle, wherein the vehicle image includes a three-dimensional scanned image of the appearance.

[0050] In this embodiment, the execution subject of the vehicle damage assessment method (for example Figure 1 The server or terminal device shown in the figure can obtain a vehicle image of the target vehicle. Specifically, the vehicle image here can include a three-dimensional scanned image of the appearance, which can three-dimensionally present the damage of the vehicle appearance, such as dents, scratches, tears, etc. In addition, the vehicle image can also include an image taken by a camera, etc. Generally speaking, the target vehicle is a vehicle to be assessed for damage, that is, the target vehicle is a damaged vehicle.

[0051] In some optional implementations of the present embodiment, the vehicle image includes an appearance shot image and an appearance three-dimensional scanning image; and obtaining a vehicle image of a target vehicle includes: obtaining a distant image of the target vehicle, performing target detection on the distant image, and obtaining damaged part information; for the damaged part indicated by the damaged part information, obtaining a close-up image and an appearance three-dimensional scanning image, and using the close-up image and the distant image as the appearance shot image; using the appearance shot image and the appearance three-dimensional scanning image as the vehicle image.

[0052] In these optional implementations, the execution subject may obtain a distant image of the target vehicle and perform target detection on the distant image, and the target detection result obtained is the damaged part information. The execution subject may obtain a close-up image and a three-dimensional scanned image of the damaged part indicated by the damaged part information. Afterwards, the execution subject may use the obtained photographed image and three-dimensional scanned image of the appearance as vehicle images, so that the execution subject may use these vehicle images to assess damage.

[0053] The damage part information here indicates the approximate damage part of the vehicle. In some cases, the specific damaged parts of the vehicle cannot be obtained. For example, when the specific damaged part of the vehicle is the left front high beam bulb of the vehicle, the damage part indicated by the damage part information can be the left front high beam bulb of the vehicle.

[0054] Specifically, if damaged part information is detected in the long-range image, the above-mentioned execution entity can send damaged part information indicating the damaged part to the camera device for the damaged part, so that the camera device switches from long-range shooting to close-up shooting, and then zooms in on the damaged part to obtain a close-up image to obtain details of the damaged part.

[0055] In some optional application scenarios of these implementations, the above-mentioned image recognition of vehicle images to obtain damaged parts information and damage degree information of the target vehicle may include: using two pre-trained semantic segmentation models to respectively process the appearance captured image and the appearance three-dimensional scanned image to obtain sub-damage information that respectively characterizes the vehicle image input into the semantic segmentation model; and obtaining damage information based on a combination of the sub-damage information respectively obtained from the two semantic segmentation models.

[0056] These application scenarios can obtain sub-damage information through appearance photographed images and appearance 3D scanning images, respectively, and thus comprehensively and accurately determine damage information through close-up images, distant images and appearance 3D scanning images.

[0057] In some optional application scenarios of these implementations, the above-mentioned target detection on the distant image to obtain the damaged part information may include: in response to the distant image containing the damaged part, performing target detection on the distant image to obtain damaged part information indicating the damaged part contained in the distant image, and position information of the damaged part in the distant image; and obtaining a close-up image and an appearance three-dimensional scanning image for the damaged part indicated by the damaged part information, including: generating close-up mode switching information based on the above-mentioned position information, and outputting a close-up mode switching instruction to the shooting device; and obtaining a close-up image of the damaged part indicated by the damaged part information.

[0058] In these optional application scenarios, when the distant image includes a damaged part, the execution subject may perform target detection on the distant image, and the detection result includes damaged part information indicating the damaged part included in the distant image, and position information of the damaged part in the distant image, such as the coordinates in the distant image. The close-up mode switching instruction here is used to instruct the camera device to switch from the distant shooting mode to the close-up shooting mode.

[0059] The execution subject may generate a close-up mode switching instruction based on the above position in a variety of ways. For example, the execution subject may generate a close-up mode switching instruction including the above position information, so that the camera device takes a photo of the corresponding damaged part on the target vehicle at the position indicated by the position information. In addition, the execution subject may also determine the shooting posture of the camera device for the damaged part based on the above position information, and generate a close-up mode switching instruction including the shooting posture.

[0060] After the execution subject outputs the close-up mode switching instruction to the camera device, the camera device can capture a close-up image of the damaged part and return it to the execution subject.

[0061] These implementations can achieve automatic and accurate acquisition of clear images of damaged parts by switching the camera to a close-up mode and capturing close-up images of damaged parts, thereby improving the accuracy of damage assessment.

[0062] Step 202, performing image recognition on the vehicle image to obtain damage information of the target vehicle, wherein the damage information is used to characterize the loss type and damage degree of the damaged parts.

[0063] In this embodiment, the above-mentioned execution entity can perform image recognition on the vehicle image, and the recognition result obtained is the damage information of the target vehicle. Specifically, the damage information can be used to characterize the damage type and damage degree of the damaged parts. Therefore, the damage information may include damaged parts information indicating the damaged parts, damage type information of the damaged parts, and damage degree information of the damaged parts. The damaged parts here are the component units of the vehicle, and are also the smallest component units that can be replaced or repaired when the vehicle is repaired, such as the left front high beam bulb, the front windshield, etc. The damage type can be dents, scratches or tears, etc. The degree of damage here can be expressed as the level of damage, the area of ​​damage and / or the length of damage, etc.

[0064] In practice, the above-mentioned execution entity can use a pre-trained neural network model for image recognition, and the neural network model can take a vehicle image of a vehicle as input and output damage information of the vehicle.

[0065] Step 203, based on the damage information, generates repair method information of the damaged parts of the target vehicle, and generates a damage assessment result including the damage information and the repair method information.

[0066] In this embodiment, the execution subject may generate repair method information of the damaged parts indicated by the damage information of the target vehicle based on the damage information, and generate a damage assessment result. Specifically, the repair method may include repairing or replacing the damaged parts of the target vehicle.

[0067] In practice, the above-mentioned execution subject may generate the repair method information based on the damage information in a variety of ways. For example, the damage degree information characterizing the damage degree of the damaged parts may correspond to n damage degree levels in total, and each damage degree level may have a corresponding repair method. For example, the repair method corresponding to the damage degree level with the lightest damage degree is to repair the damaged parts.

[0068] Specifically, the execution entity may generate the damage assessment result in a variety of ways. For example, the execution entity may directly use the damage information and the repair method information as the damage assessment result. In addition, when the damage assessment result is a damage assessment report, the execution entity may use a damage assessment report template to generate a damage assessment report including the damage information and the repair method information.

[0069] In some optional implementations of the present embodiment, repair method information of damaged parts of the target vehicle is generated based on the damage information, including: inputting the basic vehicle information and damage information of the target vehicle into a pre-trained repair method classification model, and obtaining the repair method of the damaged parts indicated by the damaged parts information output from the repair method classification model.

[0070] In these optional implementations, the above-mentioned execution subject can input the basic information of the vehicle and the damage information of the vehicle into the pre-trained maintenance method classification model, so as to obtain the maintenance method for the damaged parts output by the maintenance method classification model. The maintenance method here can be expressed as the identification of the maintenance method. For example, the identification of the maintenance method can be the identification "0" indicating "repair" or the identification "1" indicating "replacement". Specifically, the basic information of the vehicle can include at least one of the following: vehicle model, vehicle series, purchase time, and vehicle age. Different vehicle models and vehicle series may have different attributes for the same accessories, so the maintenance methods of the accessories may also be different. In addition, the purchase time and vehicle age reflect the repair value of the damaged accessories of the target vehicle itself. The repair value is inversely proportional to the vehicle age, and the earlier the purchase time, the smaller the repair value. The repair value can directly affect the maintenance method of the damaged accessories of the vehicle. For example, the greater the repair value, the greater the probability of adopting the repair method of repair.

[0071] The maintenance method classification model may be a neural network model, such as a convolutional neural network. In practice, when the damage information is text information, the maintenance method classification model is a text classification model, which can classify the damage information and determine whether the damaged parts corresponding to the damage information can be repaired or need to be replaced.

[0072] For example, the damage information is the following text information: "The front and rear pillars are pierced, and the dented and deformed part of the pillar reaches 30% of the pillar", and the repair method output by the repair method classification model can be "replacement".

[0073] These implementations can determine the repair method of damaged parts through basic vehicle information and damage information, so that the repair method can be determined more accurately according to the specific conditions of different vehicles. In addition, the accuracy of determining the repair method can be further improved through the repair method classification model.

[0074] In some optional application scenarios of these implementations, basic information of each vehicle of the target vehicle is stored in a basic information relationship graph.

[0075] In these optional application scenarios, the basic information of the vehicle can be stored in a basic information relationship graph. The basic information relationship graph of the vehicle is a basic information relationship graph composed of the basic information of each vehicle. Each node in the graph is a basic information of the target vehicle, and the node includes the attribute information of the basic information of the vehicle. For example, the model of the target vehicle is X800, and the model is used as a node in the graph. The attribute values ​​of the node include body length, displacement, engine model, etc. The attribute values ​​of the accessory node extending from the model node include accessory name, accessory price, accessory image, etc.

[0076] Through the relationship map of vehicle basic information, the basic information of each vehicle can be stored clearly and accurately, and the basic information of each vehicle can be quickly and accurately found therefrom.

[0077] In some optional implementations of the present embodiment, the above-mentioned generation of damage assessment results including damage information and repair method information may include: querying repair value reference information corresponding to the damage information; determining the repair discount rate of the target vehicle based on the vehicle additional information, and determining the repair value information corresponding to the repair value reference information based on the repair discount rate; inputting the repair value information, repair method information and vehicle basic information into a pre-generated sequence-to-sequence model to obtain text information output from the sequence-to-sequence model; and using the output text information as the damage assessment report of the target vehicle.

[0078] In these optional implementations, the above-mentioned execution entity can query the repair value reference information corresponding to the damage information. And determine the repair discount rate of the target vehicle, and then obtain the repair value information of the target vehicle. After that, the above-mentioned execution entity can use the sequence-to-sequence model to output text information to obtain a damage assessment report. Vehicle additional information is information related to the vehicle, such as the value information and / or insurance information of the vehicle. For example, the insurance information may include the insurance amount, policy number, insured type and / or insurance purchase time, etc. Different vehicle additional information may correspond to different vehicle repair discount rates.

[0079] The sequence-to-sequence model is a neural network model that can perform intelligent writing. The sequence-to-sequence model can process the information input into the model based on the damage report template, and then the above-mentioned execution entity can use the text information output from the model as the damage report.

[0080] These implementations use a sequence-to-sequence model to complete the damage assessment report, thereby realizing a fully automated damage assessment process from vehicle damage assessment to report output. The automated production of damage assessment reports increases the accuracy of the damage assessment process while improving the efficiency of damage assessment.

[0081] In some optional cases of these application scenarios, the above method may also include: obtaining an accident scene image of the target vehicle, and determining the cause of the accident corresponding to the accident scene image; and the above-mentioned inputting the repair value information, repair method information and vehicle basic information into a pre-generated sequence-to-sequence model to obtain text information output from the sequence-to-sequence model may include: inputting the cause of the accident, repair value information, and vehicle basic information into a pre-generated sequence-to-sequence model to obtain text information output from the sequence-to-sequence model.

[0082] In these optional cases, the execution subject may determine the cause of the accident corresponding to the accident scene image, and input the cause of the accident into the sequence model so that the model outputs text information including the cause of the accident.

[0083] Specifically, the execution subject may determine the cause of the accident in a variety of ways. For example, the execution subject may determine the cause of the accident only through the accident scene image, or may determine the cause of the accident through the accident scene image and other information.

[0084] In the case where the execution subject determines the cause of the accident only through the accident scene image, the execution subject may input the accident scene image into a pre-trained neural network model and obtain the cause of the accident output by the model. The neural network model may characterize the correspondence between the accident scene image and the cause of the accident presented therein.

[0085] In the case where the execution subject determines the cause of the accident through the accident scene image and other information, the execution subject may obtain the voice information of the owner of the target vehicle and convert the voice information into text, which may include the cause of the accident and may also include the license plate number, frame number, etc. of the target vehicle. The other information may include the text. The execution subject may determine the cause of the accident of the target vehicle based on the cause of the accident in the accident scene image and text. For example, the execution subject may input the cause of the accident in the accident scene image and text into a pre-trained neural network model and obtain the cause of the accident output by the model. The neural network model may characterize the correspondence between the cause of the accident in the accident scene image and text and the cause of the accident of the target vehicle. In addition, the execution subject may also send the accident scene image to the target terminal to obtain the cause of the accident returned by the target terminal, or input the neural network model to obtain the cause of the accident output by the neural network model. Afterwards, the execution subject may combine the cause of the accident with the cause of the accident in the text as the cause of the accident of the target vehicle, or, if the two are different, use the cause of the accident in the text as the cause of the accident of the target vehicle. The target terminal here may display the accident scene image to the user of the target terminal so that the user of the target terminal determines the cause of the accident.

[0086] Specifically, the repair value reference information and the repair value information can reflect the value corresponding to repairing the target vehicle with the above damage information. Generally speaking, the greater the number of damaged parts of the target vehicle, the greater the value of the damaged parts, and the greater the degree of damage, the greater the value corresponding to repairing the target vehicle.

[0087] The electronic equipment in these cases can accurately determine the cause of the accident based on the accident scene images, and intuitively reflect it in the damage assessment report.

[0088] Continue to see Figure 3 , Figure 3 FIG. 1 is a schematic diagram of an application scenario of the vehicle damage assessment method according to this embodiment. Figure 3 In the application scenario, the execution subject 301 can obtain a vehicle image 302 of the target vehicle, wherein the vehicle image includes a three-dimensional scan image of the appearance. Image recognition is performed on the vehicle image 303 to obtain damage information 303 of the target vehicle, wherein the damage information is used to characterize the loss type and damage degree of the damaged parts. Based on the damage information 303, the repair method information 304 of the damaged parts of the target vehicle is generated, and the damage assessment result 305 including the damage information 303 and the repair method information 304 is generated.

[0089] The method provided by the above-mentioned embodiment of the present application can utilize vehicle images including three-dimensional scanned images of the appearance to accurately and efficiently determine damage information including appearance damage conditions through image recognition, thereby improving the accuracy and efficiency of damage assessment.

[0090] Further references Figure 4 , which shows a process 400 of another embodiment of a vehicle damage assessment method. The process 400 of the vehicle damage assessment method includes the following steps:

[0091] Step 401, obtaining a distant image of a target vehicle, performing target detection on the distant image, and obtaining damaged part information.

[0092] In this embodiment, the execution subject (for example Figure 1 The server or terminal device shown in the figure can obtain a distant image of the target vehicle and perform target detection on the distant image. The target detection result obtained is the damaged part information. The damaged part information here indicates the approximate damaged part of the vehicle. In some cases, the parts to be repaired in the vehicle cannot be obtained. For example, when the specific damaged part of the vehicle is the left front high beam bulb of the vehicle, the damaged part indicated by the damaged part information can be the left front high beam bulb of the vehicle. Specifically, the result of the target detection can be the identification of the damaged part. The above-mentioned execution subject can use a neural network model for target detection.

[0093] Step 402: acquiring a close-up image and an appearance three-dimensional scanning image for the damaged part indicated by the damaged part information, and using the close-up image and the distant image as appearance photographed images.

[0094] In this embodiment, the execution subject may obtain a close-up image and an appearance three-dimensional scanning image of the damaged part indicated by the damaged part information. Afterwards, the execution subject may use the distant image and the close-up image as the appearance photographed image.

[0095] Specifically, if damaged part information is detected in the long-range image, the above-mentioned execution entity can send damaged part information indicating the damaged part to the camera device for the damaged part, so that the camera device switches from long-range shooting to close-up shooting, and then zooms in on the damaged part to obtain a close-up image to obtain details of the damaged part.

[0096] Step 403: Acquire a radiographic image of the target vehicle.

[0097] In this embodiment, the radiation here may be an X-ray or a gamma ray. Radiation usually has a certain penetrating ability, and the internal damage of the target vehicle can be obtained through radiation detection. Specifically, the above-mentioned execution subject can perform radiation detection on the front, rear or side of the vehicle by using the device emitting radiation, and use the obtained image as the radiation detection image.

[0098] Step 404: Use the appearance photographed image, the appearance 3D scanning image and the X-ray detection image as vehicle images.

[0099] In this embodiment, the above-mentioned execution entity may use the obtained appearance photographed images, appearance three-dimensional scanning images and X-ray detection images as vehicle images, so that the above-mentioned execution entity can use these vehicle images to assess damages.

[0100] Step 405, performing image recognition on the vehicle image to obtain damage information of the target vehicle, wherein the damage information is used to characterize the damage degree of the damaged parts.

[0101] In this embodiment, the execution subject can perform image recognition on the vehicle image, and the recognition result obtained is the damage information of the target vehicle. Specifically, the damage information can be used to characterize the damage degree of the damaged parts, and thus, the damage information can include the damaged parts information and the damage degree information of the damaged parts indicated by the damaged parts information. The damaged parts here are the components of the vehicle, and are also the smallest components that can be replaced or repaired during maintenance, such as the left front high beam bulb, the front windshield, and so on.

[0102] In some optional implementations of the present embodiment, step 405 may include: using three pre-trained semantic segmentation models to respectively process the appearance shot image, the appearance three-dimensional scanning image and the X-ray detection image to obtain sub-damage information output by each semantic segmentation model, wherein the sub-damage information includes damaged accessory information and / or damage degree information of the damaged accessory indicated by the damaged accessory information; and obtaining damage information based on a combination of the sub-damage information respectively obtained by the three semantic segmentation models.

[0103] In these optional implementations, the execution subject may use three different semantic segmentation models to process the appearance photographed image, the appearance 3D scanned image, and the X-ray detection image, respectively. For each of the three semantic segmentation models, sub-damage information corresponding to the vehicle image input into the semantic segmentation model may be obtained. The sub-damage information here may include damaged accessory information, damage type information, and / or damage degree information.

[0104] The semantic segmentation model is a neural network model (such as a deep convolutional neural network or DEEPLAB V3+, etc.) that can recognize images based on their semantics. Specifically, the output of the semantic segmentation model can be a mark indicating the location of the damage or the extent of the damage, or text information. For example, the output of the semantic segmentation model can be a mark A, which corresponds to a scratch of 10 cm.

[0105] The above-mentioned execution entity can combine the sub-damage information obtained by the three semantic segmentation models respectively, and obtain damage information based on the combination result. For example, the sub-damage information obtained by the appearance shooting image recognition is the vehicle panel, scratched 20cm. The sub-damage information obtained by the appearance three-dimensional scanning image recognition is the vehicle rear fender, bent and dented. The sub-damage information obtained by the X-ray detection image recognition is that the engine is intact. After combining the various sub-damage information, the damage information obtained is that the rear fender is bent and dented, and the panel surface is scratched 20cm.

[0106] In practice, the execution subject may adopt various methods to obtain the damage information based on the combination of each sub-damage information. For example, the execution subject may directly use the combination of each sub-damage information as the damage information. In addition, the execution subject may also convert each sub-damage information to obtain a conversion result. For example, when the sub-damage information is an identifier, the execution subject may convert the sub-damage information from the identifier to text, etc., and then combine each conversion result to obtain the damage information.

[0107] These implementations can use different semantic segmentation models to obtain rich damage information from three different images, thereby combining them to obtain a very comprehensive damage situation.

[0108] In some optional application scenarios of these implementations, the damage information is text information; and the damage information is obtained by combining the sub-damage information obtained respectively based on the three semantic segmentation models, which may include: combining the sub-damage information obtained respectively by the three semantic segmentation models into damage information to be processed; and using a pre-trained text output model to convert the damage information to be processed into damage information.

[0109] In these optional application scenarios, the above-mentioned execution subject can combine the sub-damage information generated by each semantic segmentation model, and the result of the combination is the damage information to be processed. Afterwards, the above-mentioned execution subject can input the damage information to be processed into a pre-trained text output model. Specifically, the text output model can be used to characterize the correspondence between the damage information and the text corresponding to the damage information. For example, when the damage information is an identification or text information indicating the damage situation, the text output model can be a correspondence table indicating the correspondence between the damage information and the text. In addition, if the loss information is presented in the form of an image, the text output model can be a model that can convert the image into text (such as Img2txt).

[0110] These application scenarios can convert the damage information into text information, so that the damage assessment results can be used to generate a damage assessment report that can be directly displayed to the user.

[0111] In some optional application scenarios of these implementations, the above-mentioned three pre-trained semantic segmentation models are used to process the appearance shot image, the appearance three-dimensional scanning image and the X-ray detection image respectively, including: mapping the near-field image to the distant field image to obtain a mapping result; inputting the shooting angle, damage part information and mapping result of the distant field image into the first semantic segmentation model, wherein the sub-damage information output by the first semantic segmentation model is the appearance damage quantification information; and inputting the shooting angle, damage part information and the appearance three-dimensional scanning image of the distant field image into the second semantic segmentation model, wherein the sub-damage information output by the second semantic segmentation model is the appearance three-dimensional damage information; and inputting the shooting angle, damage part information and the X-ray detection image of the distant field image into the third semantic segmentation model, wherein the sub-damage information output by the third semantic segmentation model is the internal damage information.

[0112] In these optional application scenarios, the execution subject can obtain the shooting angle of the long-range image, and input the shooting angle, the damaged part information and the vehicle image into the semantic segmentation model. In this way, the semantic segmentation model can perform image recognition on the vehicle image together with the shooting angle and the damaged part information to obtain a more accurate recognition result. The execution subject can map the near-range image to the long-range image through the alignment module, thereby obtaining a mapped vehicle image, that is, a mapping result.

[0113] The above three semantic segmentation models are trained using different training samples respectively. The training samples of the first semantic segmentation model may include mapping results and damage information annotated on the image. The training samples of the second semantic segmentation model may include an appearance three-dimensional damaged image and damage information annotated on the image. The training samples of the third semantic segmentation model may include a radiographic detection image and damage information annotated on the image. In addition, the training samples of each semantic segmentation model may also include shooting angle and damage location information.

[0114] Specifically, the first semantic segmentation model can quantify the damage to the vehicle's exterior to obtain quantitative information on the exterior damage, that is, a numerical value, such as a numerical value for the damaged area or a numerical value for the damaged length, for example, a 20 cm scratch on the board can be obtained. The second semantic segmentation model can obtain three-dimensional data of the damage, that is, three-dimensional exterior damage information, such as information indicating dents, by recognizing the exterior three-dimensional scanning image. The third semantic segmentation model can output internal damage information indicating the damage inside the vehicle by recognizing the X-ray detection image.

[0115] These application scenarios can use the shooting angle and damage location information to comprehensively identify vehicle images to obtain accurate damage information. At the same time, three different semantic segmentation models can be used to obtain quantitative information on appearance damage, appearance three-dimensional damage information, and internal damage information, thereby obtaining accurate and comprehensive vehicle damage information. In addition, by mapping the near-field image to the distant-field image, the semantic segmentation model can more accurately determine damage area, damage length, and other damage degree information.

[0116] In some optional cases of these application scenarios, the above-mentioned shooting angle can be obtained by the following steps: using a pre-trained angle classification model to classify the shooting angle of the distant view image to obtain the shooting angle of the distant view image.

[0117] In these optional cases, the execution subject uses the classification model to classify the shooting angle of the distant image, and obtains the shooting angle output from the classification model. The angle classification model here can be a neural network capable of classification, such as a residual neural network or a support vector machine model, etc. Specifically, the shooting angle can be divided into multiple categories, such as 360 categories, and the result output by the angle classification model is one of them.

[0118] These optional situations use classification models to determine the shooting angle, which can obtain more accurate shooting angles.

[0119] Step 406: Generate maintenance method information of the target vehicle based on the damage information.

[0120] In this embodiment, the execution subject may generate maintenance method information of the target vehicle based on the damage information. The maintenance method information may represent the maintenance method of the vehicle. Specifically, the maintenance method may include repairing or replacing damaged parts of the vehicle.

[0121] Further references Figure 5 As an implementation of the methods shown in the above figures, the present application provides an embodiment of a vehicle damage assessment device, which is similar to Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0122] like Figure 5 As shown, the vehicle damage assessment device 500 of this embodiment includes: an acquisition unit 501, an identification unit 502 and a generation unit 503. The acquisition unit 501 is configured to acquire a vehicle image of a target vehicle, wherein the vehicle image includes a three-dimensional scanned image of the appearance; the identification unit 502 is configured to perform image recognition on the vehicle image to obtain damage information of the target vehicle, wherein the damage information is used to characterize the loss type and damage degree of the damaged parts; the generation unit 503 is configured to generate repair method information of the damaged parts of the target vehicle based on the damage information, and generate a damage assessment result including the damage information and the repair method information.

[0123] In some embodiments, the acquisition unit 501 of the vehicle damage assessment device 500 can acquire a vehicle image of the target vehicle. Specifically, the vehicle image here can include a three-dimensional scanned image of the appearance, which can three-dimensionally present the damage of the vehicle appearance, such as dents, scratches, tears, etc. In addition, the vehicle image can also include an image captured by a camera, etc. Generally speaking, the target vehicle is a vehicle to be assessed for damage, that is, the target vehicle is a damaged vehicle.

[0124] In some embodiments, the recognition unit 502 may perform image recognition on the vehicle image, and the recognition result obtained is the damage information of the target vehicle. Specifically, the damage information may be used to characterize the damage type and damage degree of the damaged parts, and thus, the damage information may include damaged parts information indicating the damaged parts, damage type information of the damaged parts, and damage degree information of the damaged parts.

[0125] In some embodiments, the generation unit 503 may generate repair method information of the damaged parts indicated by the damage information of the target vehicle based on the damage information, and generate a damage assessment result. Specifically, the repair method may include repairing or replacing the damaged parts of the target vehicle.

[0126] In some optional implementations of this embodiment, the vehicle image includes an appearance shot image, an appearance three-dimensional scanning image and a radiographic detection image; and an acquisition unit, including: an acquisition module, configured to acquire a distant image of the target vehicle, perform target detection on the distant image, and obtain damaged part information; a first image module, configured to acquire a close-up image and an appearance three-dimensional scanning image for the damaged part indicated by the damaged part information, and use the close-up image and the distant image as the appearance shot image; a second image module, configured to acquire a radiographic detection image of the target vehicle; and a third image module, configured to use the appearance shot image, the appearance three-dimensional scanning image and the radiographic detection image as the vehicle image.

[0127] In some optional implementations of the present embodiment, the identification unit includes: a separate processing module, configured to use three pre-trained semantic segmentation models to respectively process the appearance shot image, the appearance three-dimensional scanning image and the X-ray detection image, to obtain sub-damage information output by each semantic segmentation model, wherein the sub-damage information includes at least one of the following: damaged accessory information, damage type information of the damaged accessory indicated by the damaged accessory information, and damage degree information of the damaged accessory; a combination module, configured to obtain damage information based on a combination of the sub-damage information respectively obtained by the three semantic segmentation models.

[0128] In some optional implementations of the present embodiment, the respective processing modules are further configured to: map the near-field image to the distant-field image to obtain a mapping result; input the shooting angle, damage part information and mapping result of the distant-field image into a first semantic segmentation model, wherein the sub-damage information output by the first semantic segmentation model is appearance damage quantification information; input the shooting angle, damage part information and appearance three-dimensional scanning image of the distant-field image into a second semantic segmentation model, wherein the sub-damage information output by the second semantic segmentation model is appearance three-dimensional damage information; input the shooting angle, damage part information and X-ray detection image of the distant-field image into a third semantic segmentation model, wherein the sub-damage information output by the third semantic segmentation model is internal damage information.

[0129] In some optional implementations of this embodiment, the vehicle image includes an appearance shot image and an appearance three-dimensional scanning image; and obtaining a vehicle image of a target vehicle includes: a detection module, configured to obtain a distant image of the target vehicle, perform target detection on the distant image, and obtain damaged part information; a fourth image module, configured to obtain a close-up image and an appearance three-dimensional scanning image for the damaged part indicated by the damaged part information, and use the close-up image and the distant image as the appearance shot image; and a fifth image module, configured to use the appearance shot image and the appearance three-dimensional scanning image as the vehicle image.

[0130] In some optional implementations of this embodiment, the acquisition module or the detection module is further configured to: in response to the distant image containing a damaged part, perform target detection on the distant image to obtain damaged part information indicating the damaged part contained in the distant image, and position information of the damaged part in the distant image; and the first image module is further configured to: generate a close-up mode switching instruction based on the position information, and output the close-up mode switching instruction to the shooting device; and obtain a close-up image of the damaged part indicated by the damaged part information.

[0131] In some optional implementations of this embodiment, the damage information is text information; and the combination module is further configured to: combine the sub-damage information obtained by the three semantic segmentation models into damage information to be processed; and use a pre-trained text output model to convert the damage information to be processed into damage information.

[0132] In some optional implementations of this embodiment, the generation unit includes: a method determination module, which is configured to input the basic vehicle information and damage information of the target vehicle into a pre-trained repair method classification model, and obtain the repair method for the damaged parts indicated by the damaged parts information output from the repair method classification model.

[0133] In some optional implementations of the present embodiment, the generation unit includes: a query module, configured to query the repair value reference information corresponding to the damage information; a value determination module, configured to determine the repair discount rate of the target vehicle based on the vehicle additional information, and determine the repair value information corresponding to the repair value reference information based on the repair discount rate; a text output module, configured to input the repair value information, repair method information and vehicle basic information into a pre-generated sequence-to-sequence model to obtain text information output from the sequence-to-sequence model; and a report generation module, configured to use the output text information as a damage assessment report of the target vehicle.

[0134] In some optional implementations of this embodiment, the device also includes: a cause determination unit, configured to obtain an accident scene image of the target vehicle and determine the cause of the accident corresponding to the accident scene image; and a text output module, further configured to: input the cause of the accident, repair value information, repair method information and vehicle basic information into a pre-generated sequence-to-sequence model to obtain text information output from the sequence-to-sequence model.

[0135] In some optional implementations of this embodiment, the shooting angle is acquired through the following steps: using a pre-trained angle classification model to classify the shooting angle of the distant view image to obtain the shooting angle of the distant view image.

[0136] In some optional implementations of this embodiment, basic information of each vehicle of the target vehicle is stored in a basic information relationship map.

[0137] like Figure 6 As shown, the electronic device 600 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device 600 are also stored. The processing device 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0138] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device 600 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 6 The electronic device 600 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead. Figure 6 Each block shown in the figure may represent one device, or may represent multiple devices as required.

[0139] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through a communication device 609, or installed from a storage device 608, or installed from a ROM 602. When the computer program is executed by the processing device 601, the above functions defined in the method of the embodiment of the present disclosure are executed. It should be noted that the computer-readable medium of the embodiment of the present disclosure can be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium can be, for example, - but not limited to - an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection with one or more conductors, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In an embodiment of the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, an apparatus, or a device. In an embodiment of the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in combination with an instruction execution system, an apparatus, or a device. The program code embodied on the computer readable medium may be transmitted using any appropriate medium, including but not limited to: wire, optical cable, RF (radio frequency), etc., or any suitable combination of the foregoing.

[0140] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0141] The units involved in the embodiments described in the present application may be implemented by software or hardware. The units described may also be set in a processor, for example, it may be described as: a processor includes an acquisition unit, an identification unit, and a generation unit. The names of these units do not constitute a limitation on the units themselves in some cases, for example, the acquisition unit may also be described as a "unit for acquiring a vehicle image of a target vehicle".

[0142] As another aspect, the present application also provides a computer-readable medium, which may be included in the device described in the above embodiment; or it may exist independently without being assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by the device, the device: obtains a vehicle image of the target vehicle, wherein the vehicle image includes a three-dimensional scan image of the appearance; performs image recognition on the vehicle image to obtain damage information of the target vehicle, wherein the damage information is used to characterize the loss type and degree of damage of the damaged parts; based on the damage information, generates repair method information of the damaged parts of the target vehicle, and generates a damage assessment result including the damage information and the repair method information.

[0143] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above invention concept. For example, the above features are replaced with the technical features with similar functions disclosed in this application (but not limited to) by each other to form a technical solution.

Claims

1. A vehicle damage assessment method, the method comprising: Acquiring a vehicle image of a target vehicle, including: acquiring a distant image of the target vehicle, performing target detection on the distant image to obtain damaged part information, acquiring a close-up image and an appearance three-dimensional scanning image for the damaged part indicated by the damaged part information, and using the close-up image and the distant image as the appearance photographed image; acquiring a radiographic detection image of the target vehicle; and using the appearance photographed image, the appearance three-dimensional scanning image, and the radiographic detection image as the vehicle image; Performing image recognition on the vehicle image to obtain damage information of the target vehicle, including: mapping the near-view image to the distant view image to obtain a mapping result; inputting the shooting angle of the distant view image, the damage part information and the mapping result into a first semantic segmentation model, wherein the sub-damage information output by the first semantic segmentation model is appearance damage quantification information; inputting the shooting angle of the distant view image, the damage part information and the appearance three-dimensional scanning image into a second semantic segmentation model, wherein the sub-damage information output by the second semantic segmentation model is appearance three-dimensional damage information; inputting the shooting angle of the distant view image, the damage part information and the radiographic detection image into a third semantic segmentation model, wherein the sub-damage information output by the third semantic segmentation model is internal damage information; obtaining the damage information based on a combination of the sub-damage information respectively obtained by the three semantic segmentation models, wherein the damage information is used to characterize the loss type and degree of damage of the damaged accessories; Based on the damage information, repair method information of the damaged parts of the target vehicle is generated, and a damage assessment result including the damage information and the repair method information is generated.

2. The method according to claim 1, wherein: The shooting angle is obtained by the following steps: The shooting angle of the distant view image is classified by using a pre-trained angle classification model to obtain the shooting angle of the distant view image.

3. The method according to claim 1, wherein: The vehicle image includes an appearance photographed image and an appearance three-dimensional scanned image; as well as The step of acquiring a vehicle image of a target vehicle comprises: Acquire a distant image of the target vehicle, perform target detection on the distant image, and obtain damaged part information; For the damaged part indicated by the damaged part information, a close-up image and an appearance three-dimensional scanning image are acquired, and the close-up image and the distant image are used as appearance photographed images; The appearance photographed image and the appearance three-dimensional scanned image are used as the vehicle image.

4. The method according to claim 1 or 3, wherein: The performing target detection on the distant image to obtain damaged part information includes: In response to the distant image containing a damaged part, performing target detection on the distant image to obtain damaged part information indicating the damaged part contained in the distant image and position information of the damaged part in the distant image; and The step of acquiring a close-up image and an appearance three-dimensional scanning image for the damaged part indicated by the damaged part information includes: Based on the position information, generating a close-up mode switching instruction, and outputting the close-up mode switching instruction to a photographing device; A close-up image of the damaged part indicated by the damaged part information is obtained.

5. The method according to claim 1, wherein: The damage information is text information; as well as The damage information is obtained by combining the sub-damage information respectively obtained based on the three semantic segmentation models, including: Combining the sub-damage information obtained by the three semantic segmentation models into damage information to be processed; The to-be-processed damage information is converted into the damage information by using a pre-trained text output model.

6. The method according to claim 1, wherein: Generating the repair method information of the damaged parts of the target vehicle based on the damage information includes: The basic vehicle information and the damage information of the target vehicle are input into a pre-trained maintenance method classification model to obtain a maintenance method for the damaged parts indicated by the damaged parts information output from the maintenance method classification model.

7. The method according to claim 1, wherein: The generating of the damage assessment result including the damage information and the repair method information includes: Querying the repair value reference information corresponding to the damage information; Determining a maintenance discount rate of the target vehicle based on the vehicle additional information, and determining maintenance value information corresponding to the maintenance value reference information based on the maintenance discount rate; Inputting the maintenance value information, the maintenance method information and the basic vehicle information into a pre-generated sequence-to-sequence model to obtain text information output from the sequence-to-sequence model; The output text information is used as the damage assessment report of the target vehicle.

8. The method according to claim 7, wherein: The method further comprises: Acquire an accident scene image of the target vehicle, and determine the cause of the accident corresponding to the accident scene image; and The step of inputting the maintenance value information, the maintenance method information and the basic vehicle information into a pre-generated sequence-to-sequence model to obtain text information output from the sequence-to-sequence model includes: The cause of the accident, the repair value information, the repair method information and the basic vehicle information are input into a pre-generated sequence-to-sequence model to obtain text information output from the sequence-to-sequence model.

9. The method according to claim 6, wherein: The basic information of each vehicle of the target vehicle is stored in a basic information relationship map.

10. A vehicle damage assessment device, comprising: The acquisition unit is configured to acquire a vehicle image of a target vehicle, including: an acquisition module, configured to acquire a distant image of the target vehicle, perform target detection on the distant image, and obtain damaged part information; a first image module, configured to acquire a close-up image and an appearance three-dimensional scanning image for the damaged part indicated by the damaged part information, and use the close-up image and the distant image as the appearance photographed image; a second image module, configured to acquire a radiographic detection image of the target vehicle; and a third image module, configured to use the appearance photographed image, the appearance three-dimensional scanning image, and the radiographic detection image as the vehicle image; The recognition unit is configured to perform image recognition on the vehicle image to obtain damage information of the target vehicle, including: a separate processing module, configured to map the near-view image to the distant view image to obtain a mapping result; input the shooting angle of the distant view image, the damage part information and the mapping result into a first semantic segmentation model, wherein the sub-damage information output by the first semantic segmentation model is appearance damage quantification information; input the shooting angle of the distant view image, the damage part information and the appearance three-dimensional scanning image into a second semantic segmentation model, wherein the sub-damage information output by the second semantic segmentation model is appearance three-dimensional damage information; input the shooting angle of the distant view image, the damage part information and the radiographic detection image into a third semantic segmentation model, wherein the sub-damage information output by the third semantic segmentation model is internal damage information; and a combination module, configured to obtain the damage information based on a combination of the sub-damage information respectively obtained by the three semantic segmentation models, wherein the damage information is used to characterize the loss type and degree of damage of the damaged accessories; The generating unit is configured to generate the repair method information of the damaged parts of the target vehicle based on the damage information, and to generate a damage assessment result including the damage information and the repair method information.

11. The device according to claim 10, wherein: The acquisition module is further configured to: In response to the distant image containing a damaged part, performing target detection on the distant image to obtain damaged part information indicating the damaged part contained in the distant image and position information of the damaged part in the distant image; and The first image module is further configured to: Based on the position information, generating a close-up mode switching instruction, and outputting the close-up mode switching instruction to a photographing device; A close-up image of the damaged part indicated by the damaged part information is obtained.

12. The device according to claim 10, wherein: The damaged information is text information; and The combined module is further configured to: Combining the sub-damage information obtained by the three semantic segmentation models into damage information to be processed; The to-be-processed damage information is converted into the damage information by using a pre-trained text output model.

13. The device according to claim 10, wherein: The generating unit comprises: The method determination module is configured to input the basic vehicle information and the damage information of the target vehicle into a pre-trained maintenance method classification model, and obtain the maintenance method for the damaged parts indicated by the damaged parts information output from the maintenance method classification model.

14. The device according to claim 10, wherein: The generating unit comprises: A query module configured to query repair value reference information corresponding to the damage information; a value determination module configured to determine a maintenance discount rate of the target vehicle based on the vehicle additional information, and to determine maintenance value information corresponding to the maintenance value reference information based on the maintenance discount rate; a text output module configured to input the maintenance value information, the maintenance method information and the basic vehicle information into a pre-generated sequence-to-sequence model to obtain text information output from the sequence-to-sequence model; The report generation module is configured to use the output text information as the damage assessment report of the target vehicle.

15. The device according to claim 14, wherein: The device also includes: a cause determination unit configured to obtain an accident scene image of the target vehicle and determine an accident cause corresponding to the accident scene image; and The text output module is further configured to: The cause of the accident, the repair value information, the repair method information and the basic vehicle information are input into a pre-generated sequence-to-sequence model to obtain text information output from the sequence-to-sequence model.

16. An electronic device, comprising: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 9.

17. A computer-readable storage medium having a computer program stored thereon, wherein: When the program is executed by a processor, the method according to any one of claims 1 to 9 is implemented.

Images