WO2019205376A1 - Vehicle damage determination method, server, and storage medium - Google Patents
Vehicle damage determination method, server, and storage medium Download PDFInfo
- Publication number
- WO2019205376A1 WO2019205376A1 PCT/CN2018/102121 CN2018102121W WO2019205376A1 WO 2019205376 A1 WO2019205376 A1 WO 2019205376A1 CN 2018102121 W CN2018102121 W CN 2018102121W WO 2019205376 A1 WO2019205376 A1 WO 2019205376A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- determined
- vehicle
- photo
- damage
- preset
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
Definitions
- the present application relates to the field of picture recognition technologies, and in particular, to a vehicle loss determination method, a server, and a computer readable storage medium.
- the present application provides a vehicle damage determination method, a server, and a computer readable storage medium, the main purpose of which is to improve the comprehensiveness and accuracy of detection of damaged parts of a vehicle.
- the present application provides a vehicle damage determination method, including:
- Receiving step receiving a loss request and a photo to be determined by the user to be determined;
- Classification step analyzing the to-be-determined loss photo by using a pre-trained classification model, and obtaining classification information of each part of the vehicle in each photo to be determined;
- Angle determining step determining, according to the classification information of each part of the vehicle in each photo to be determined, using a predetermined shooting angle determination rule to determine the shooting angle of each photo to be determined;
- Determining the damage Combining the shooting angle and the classification information of the vehicle part, using the pre-trained fixed loss model to analyze the damage of each part of the vehicle in each photo to be determined, and output the damage analysis result of the vehicle in each photo to be determined.
- the present application further provides a server, the server includes: a memory and a processor, wherein the memory stores a vehicle damage determination program, where the vehicle damage determination program is executed by the processor, and the following steps can be implemented:
- Receiving step receiving a loss request and a photo to be determined by the user to be determined;
- Classification step analyzing the to-be-determined loss photo by using a pre-trained classification model, and obtaining classification information of each part of the vehicle in each photo to be determined;
- the angle determining step determining, according to the classification information of each part of the vehicle in each photo to be determined, the shooting angle of each photo to be determined by using a predetermined shooting angle determination rule;
- Determining the damage Combining the shooting angle and the classification information of the vehicle part, using the pre-trained fixed loss model to analyze the damage of each part of the vehicle in each photo to be determined, and output the damage analysis result of the vehicle in each photo to be determined.
- the present application further provides a computer readable storage medium including a vehicle damage determination program, where the vehicle damage determination program is executed by a processor, as described above Any step in the vehicle damage determination method.
- the vehicle damage determination method, the server and the computer readable storage medium proposed by the present application classify each part of the vehicle in the to-be-determined loss picture uploaded by the user by using the classification model, and then determine the shooting angle of the photo by using a predetermined shooting angle determination rule. Finally, the classification information and shooting angle of the vehicle parts are analyzed by the damage model to analyze the damage of the vehicle parts, and the damage analysis results are fed back to comprehensively test the vehicle to improve the detection accuracy.
- FIG. 1 is a schematic diagram of a preferred embodiment of a server of the present application.
- FIG. 2 is a block diagram showing a preferred embodiment of the vehicle damage determination program of FIG. 1;
- FIG. 3 is a flow chart of a preferred embodiment of a method for determining a vehicle damage according to the present application
- FIG. 5 is a flow chart of the training of the fixed loss model of the present application.
- FIG. 1 it is a schematic diagram of a preferred embodiment of the server 1 of the present application.
- the server 1 refers to a car insurance claim server, which may be a server, a smart phone, a tablet computer, a personal computer, a portable computer, and other electronic devices having computing functions.
- the server 1 includes a memory 11, a processor 12, a network interface 13, and a communication bus 14.
- the network interface 13 can optionally include a standard wired interface and a wireless interface (such as a WI-FI interface).
- Communication bus 14 is used to implement connection communication between these components.
- the memory 11 includes at least one type of readable storage medium.
- the at least one type of readable storage medium may be a non-volatile storage medium such as a flash memory, a hard disk, a multimedia card, a card type memory, or the like.
- the memory 11 may be an internal storage unit of the server 1, such as a hard disk of the server 1.
- the memory 11 may also be an external storage unit of the server 1, such as a plug-in hard disk equipped on the server 1, a smart memory card (SMC), and a secure digital ( Secure Digital, SD) cards, flash cards, etc.
- SMC smart memory card
- SD Secure Digital
- the memory 11 can be used not only for storing application software and various types of data installed in the server 1, such as the vehicle damage determination program 10 and the like.
- the processor 12 in some embodiments, may be a Central Processing Unit (CPU), microprocessor or other data processing chip for running program code or processing data stored in the memory 11, such as performing vehicle damage determination.
- CPU Central Processing Unit
- microprocessor or other data processing chip for running program code or processing data stored in the memory 11, such as performing vehicle damage determination.
- the computer program code of the program 10 the classification model, and the training of the fixed loss model.
- FIG. 1 shows only server 1 with components 11-14 and vehicle damage determination program 10, but it should be understood that not all illustrated components may be implemented and that more or fewer components may be implemented instead.
- the server 1 may also include a display, which may be referred to as a display screen or a display unit.
- the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an Organic Light-Emitting Diode (OLED) touch sensor.
- the display is used to display information processed in the server 1 and a work interface for displaying visualizations, such as showing the extent of damage to various parts of the vehicle.
- the server 1 may further include a user interface
- the user interface may include an input unit such as a keyboard, a voice output device such as an audio, a headset, etc.
- the user interface may further include a standard wired interface and a wireless interface.
- the server 1 may also include radio frequency (RF) circuits, sensors, audio circuits, and the like, and details are not described herein.
- RF radio frequency
- FIG. 2 is a block diagram of a preferred embodiment of the vehicle damage determination program 10 of FIG.
- a module as referred to in this application refers to a series of computer program instructions that are capable of performing a particular function.
- the vehicle damage determination program 10 includes: a receiving module 110, a classification module 120, an angle determining module 130, and a loss detecting module 140.
- the functions or operating steps implemented by the modules 110-140 are similar to the above. No more details here, exemplarily, for example:
- the receiving module 110 is configured to receive the loss request and the to-be-determined loss picture uploaded by the user.
- the fixed loss request and the to-be-determined loss photo of the vehicle are sent by the user to the server 1 through the handheld terminal.
- the photo to be determined refers to a photo of the panorama of the vehicle to be damaged.
- the user uploads a photo of the panoramic view of the vehicle to be damaged using the mobile insurance claim application APP of the mobile phone, and initiates a self-service insurance claim.
- the classification module 120 is configured to analyze the to-be-determined photo by using a pre-trained classification model, and obtain classification information of each part of the vehicle in each photo to be determined.
- the classification information of each part of the vehicle includes: a left front door, a left front fender, a left front window, a left front light, a left rear door, a left rear fender, a left rear window, a left rear light, a right front door, a right front Leaf panel, right front window, right front light, right rear door, right rear fender, right rear window, right rear light, front window, rear window, front license plate, rear license plate and vehicle identification code.
- the pre-trained classification model is an SSD model, and the specific structure of the classification model is as shown in Table 1:
- the Layer Name column indicates the name of each layer
- Input indicates the input layer
- Conv indicates the convolution layer of the model
- Conv1 indicates the first convolution layer of the model
- MaxPool indicates the maximum pooling layer of the model
- MaxPool1 indicates the model.
- the first maximum pooling layer, Fc represents the fully connected layer in the model, Fc1 represents the first fully connected layer in the model, Softmax represents the Softmax classifier; Batch Size represents the number of input images of the current layer; Kernel Size represents the current layer The scale of the convolution kernel (for example, Kernel Size can be equal to 3, indicating that the scale of the convolution kernel is 3*3); Stride Size indicates the moving step size of the convolution kernel, that is, moving to the next convolution after completing one convolution The distance of the position; Pad Size indicates the size of the image fill in the current network layer; norm represents the layer obtained by normalizing the points on the feature map; mbox_loc is used to predict the regression value of the bounding box; mbox_conf is used to predict the feature The category of each point on the map; mbox_priorbox is used to generate the bounding box; mbox_loss is used to calculate the loss function of the bounding box.
- the angle determining module 130 is configured to determine, according to the classification information of each part of the vehicle in each photo to be determined, the shooting angle of each photo to be determined by using a predetermined shooting angle determination rule.
- the predetermined shooting angle determination rule includes:
- a part of the vehicle to be determined to be determined is a left front door, a left front fender, a left front window, and a left front light, it is determined that the shooting angle of the to-be-determined photo is a first preset angle;
- the shooting angle of the to-be-determined photo is determined to be a second preset angle
- a part of the vehicle to be analyzed is a right front door, a right front fender, a right front window, and a right front light, it is determined that the shooting angle of the to-be-determined photo is a third preset angle;
- the shooting angle of the to-be-determined photo is determined to be a fourth preset angle.
- the preset angle includes a first preset angle, a second preset angle, a third preset angle, and a fourth preset angle.
- the preset angle includes 45 degrees to the front, 45 degrees to the left, 45 degrees to the front, and 45 degrees to the right. Further, if it is analyzed that the shooting angle does not include a certain preset angle, the server 1 sends a prompt message to the user handheld terminal: if the **45 degree pending loss photo is missing, please upload it in time.
- the damage module 140 is configured to analyze the damage situation of each part of the vehicle by using the pre-trained fixed loss model in combination with the shooting angle and the classification information of the vehicle part, and output the damage analysis result of the vehicle in each photo to be determined.
- the damage situation analysis is performed for each vehicle part by using the corresponding fixed loss model.
- the training is performed using the fixed loss model of the left front door.
- the corresponding fixed loss model is pre-trained, and the fixed loss model is a VGG-16 model, and the specific structure of the fixed loss model is as shown in Table 2:
- Layer Name column indicates the name of each layer
- Input indicates the input layer
- Conv indicates the convolution layer of the model
- Conv1 indicates the first convolution layer of the model
- MaxPool indicates the maximum pooling layer of the model
- MaxPool1 indicates the model.
- the first maximum pooling layer, Fc represents the fully connected layer in the model
- Fc1 represents the first fully connected layer in the model
- Softmax represents the Softmax classifier
- Batch Size represents the number of input images of the current layer
- Kernel Size represents the current layer
- the scale of the convolution kernel (for example, Kernel Size can be equal to 3, indicating that the scale of the convolution kernel is 3*3)
- Stride Size indicates the moving step size of the convolution kernel, that is, moving to the next convolution after completing one convolution The distance of the location
- Pad Size indicates the size of the image fill in the current network layer.
- FIG. 3 it is a flowchart of a preferred embodiment of the vehicle damage determination method of the present application.
- the method for implementing the vehicle damage determination when the processor 12 executes the computer program of the vehicle damage determination program 10 stored in the memory 11 includes: Step S10 - Step S40:
- the receiving module 110 receives the fixed loss request sent by the user and the to-be-determined loss picture of the uploaded vehicle.
- the user can use the mobile phone to take a picture of the vehicle's panoramic car damage at the scene of the accident, and upload the photo to the auto insurance claim application APP to initiate the self-service insurance claim.
- the vehicle damage photo captures the panoramic view of the vehicle from a preset angle to prevent missing detection of the damaged portion.
- the preset angle includes a first preset angle, a second preset angle, a third preset angle, and a fourth preset angle.
- the first preset angle, the second preset angle, the third preset angle, and the fourth preset angle respectively correspond to a left front 45 degrees, a left rear 45 degrees, a right front 45 degrees, and a right rear 45 degrees.
- step S20 the classification module 120 analyzes the to-be-determined loss photos uploaded by the user by using the pre-trained classification model, and acquires classification information of each part of the vehicle in each photo to be determined.
- the classification model is an SSD model, and the classification model is pre-trained.
- FIG. 4 it is a flowchart of the training of the classification model of the present application, and the training steps of the model are as follows:
- the classification labeling means that different vehicle parts are respectively framed by using frame lines of different colors, and each frame line area is classified and labeled.
- the frame line of each color corresponds to a part area.
- the green frame line corresponds to the front license plate area
- the red frame line corresponds to the vehicle identification code area.
- the classified sample picture is divided into a training set of a first ratio and a verification set of a second ratio, wherein the first ratio is greater than the second ratio. For example, 80% of the sample pictures, that is, 80,000 labeled sample pictures are randomly used as the training set, and the remaining 20% of the sample pictures, that is, 20,000 labeled sample pictures, are used as the verification set.
- the classification model is trained by using the sample picture in the training set, and the accuracy of the classification model is verified by using the sample picture in the verification set after the training. For example, 80,000 sample images in the training set are input into the SSD model for training, a classification model is generated, and 20,000 sample images in the verification set are input into the generated classification model to verify the accuracy of the model.
- the training is completed. If the accuracy is less than or equal to the preset value, the number of sample pictures is increased, and then the steps of dividing the sample picture into the training set and the verification set are returned. Assume that the default value is 96%. If the verification accuracy is greater than 96%, the training is completed. If the accuracy is less than or equal to 96%, add 20,000 sample images after classification, and then return to divide the sample image into training sets. And the steps of the validation set.
- step S30 the angle determining module 130 determines the shooting angle of each photo to be determined by using the predetermined shooting angle determination rule according to the classification information of each part of the vehicle in the photo to be determined.
- the predetermined shooting angle determination rule includes:
- the shooting angle of the to-be-determined photo is the first preset angle. For example, if a photo contains a left front door, a left front fender, a left front window, and a left front light, it is judged that the photographing angle of the photograph is 45 degrees to the left.
- a part of the vehicle to be analyzed is a left rear door, a left rear fender, a left rear window, and a left rear light
- the shooting angle of the to-be-determined photo is a second preset angle. For example, if a photo includes a left rear door, a left rear fender, a left rear window, and a left rear light, it is determined that the photographing angle of the photograph is 45 degrees to the left.
- a part of the vehicle to be analyzed is a right front door, a right front fender, a right front window, and a right front light
- the shooting angle of the to-be-determined photo is a third preset angle. For example, if a photo includes a right front door, a right front fender, a right front window, and a right front light, it is determined that the photographing angle of the photograph is 45 degrees to the front.
- the shooting angle of the to-be-determined photo is determined to be a fourth preset angle. For example, if a photo includes a right rear door, a right rear fender, a right rear window, and a right rear light, it is determined that the photographing angle of the photograph is 45 degrees to the right.
- the server 1 sends the prompt information to the user handheld terminal. For example, if the photo uploaded by the user does not include the photos of the right rear door, the right rear fender, the right rear window, and the right rear light, the angle determination module 130 sends a prompt message to the user's mobile phone: the missing right 45 degrees is to be determined. Please upload the photo in time.
- the fixed loss module 140 combines the shooting angle and the classification information of the vehicle part, analyzes the damage situation of each part of the vehicle in each photo to be determined by using the pre-trained fixed loss model, and outputs the damage analysis of the vehicle in each photo to be determined. result.
- the combined shooting angle and the classification information of the vehicle part are used to detect the damaged part by calling the corresponding fixed loss model on each part of the vehicle, and determining the damaged area, the damage degree and the compensation amount of the damaged part according to the shooting angle.
- the user is prompted in a preset manner. For example, the user is prompted by SMS to damage the various parts of the vehicle and the amount of compensation.
- the fixed loss model is a VGG-16 model, and the fixed loss model is pre-trained. As shown in FIG. 5, it is a flowchart of the training of the fixed loss model of the present application, and the training steps of the model are as follows:
- the portion of the scale has a sample of the damaged sample and a sample of the sample at the fourth ratio where there is no damage. Assuming that the third ratio is 60% and the fourth ratio is 40%, 100,000 sample images of the part contain 60,000 samples of the damaged part of the part and 40,000 sample pictures of the part without damage.
- different levels of damage may be set.
- the damage level of the sample image needs to be marked, and the damage level includes serious damage, serious damage, and slight damage.
- a certain proportion is set for three different damage levels in the sample picture with damage. For example, severe damage accounts for 40% of the sample images with damage, serious damage accounts for 30% of the damaged sample images, and severe damage accounts for 30% of the damaged sample images.
- the sample picture of the part after the damage is marked is randomly divided into a training set of a fifth ratio and a verification set of a sixth ratio, wherein the fifth ratio is greater than the sixth ratio.
- the sample picture after 70% damage is randomly selected, that is, the 70,000-labeled sample picture is used as the training set, and the remaining 30% of the damaged sample picture, that is, the 30,000-labeled sample picture is used as the verification set.
- the fixed loss model is trained by using the sample picture in the training set, and the accuracy of the fixed loss model is verified by using the sample picture in the verification set after the training. For example, the sample images of the 70,000 damages in the training set are input into the SSD model for training, and a fixed loss model is generated, and the sample images of the 20,000 lesions in the verification set are input into the generated fixed loss model to verify the model. The accuracy rate.
- the training is completed. If the accuracy is less than or equal to the preset threshold, the number of sample pictures of the vehicle part is increased, and then the step of dividing the sample picture into the training set and the verification set is returned. Assume that the preset threshold is 98%. If the verification accuracy is greater than 98%, the training is completed. If the accuracy is less than or equal to 98%, then 20,000 samples of the damaged label are added, and then the sample is divided into The steps of the training set and the validation set.
- the vehicle damage determination method is configured to receive the vehicle damage photos of the uploaded vehicle panorama and classify the vehicle damage photos by using the classification model, identify each part of the vehicle, and then determine the shooting angle according to the vehicle part information in the photograph, and combine The shooting angle and the vehicle part information are used to perform the loss analysis using the corresponding fixed loss model, and the damage analysis result is output, and the analysis error rate is reduced to prevent the missing part from being detected.
- the embodiment of the present application further provides a computer readable storage medium, where the computer readable storage medium includes a vehicle damage determination program 10, and when the vehicle damage determination program 10 is executed by the processor, the following operations are implemented:
- Receiving step receiving a loss request and a photo to be determined by the user to be determined;
- Classification step analyzing the to-be-determined loss photo by using a pre-trained classification model, and obtaining classification information of each part of the vehicle in each photo to be determined;
- the angle determining step determining, according to the classification information of each part of the vehicle in each photo to be determined, the shooting angle of each photo to be determined by using a predetermined shooting angle determination rule;
- Determining the damage Combining the shooting angle and the classification information of the vehicle part, using the pre-trained fixed loss model to analyze the damage of each part of the vehicle in each photo to be determined, and output the damage analysis result of the vehicle in each photo to be determined.
- the classification model is an SSD model
- the training steps of the classification model are as follows:
- the classification model is trained by using the sample picture in the training set, and the accuracy of the classification model is verified by using the sample picture in the verification set after the training;
- the training is completed. If the accuracy is less than or equal to the preset value, the number of sample pictures is increased, and then the steps of dividing the sample picture into the training set and the verification set are returned.
- the fixed loss model is a VGG-16 model
- the training steps of the fixed loss model are as follows:
- the second preset number of sample pictures includes the third ratio of the a sample picture of the damaged part at the site and a sample picture of the fourth ratio where the part is not damaged;
- the sample picture of the part after the damage label is randomly divided into a training set of a fifth ratio and a verification set of a sixth ratio, wherein the fifth ratio is greater than the sixth ratio;
- the determined loss model is trained by using the sample picture in the training set, and the accuracy of the fixed loss model is verified by using the sample picture in the verification set after the training;
- the training is completed. If the accuracy is less than or equal to the preset threshold, the number of sample pictures of the part is increased, and then the step of dividing the sample picture into the training set and the verification set is returned.
- the method further comprises:
- Prompt step If the determined shooting angle does not include a preset angle, the user is reminded to re-shoot and upload the to-be-determined photo of the preset angle.
- the preset angle includes a first preset angle, a second preset angle, a third preset angle, and a fourth preset angle
- the predetermined shooting angle determination rule includes:
- a part of the vehicle to be determined to be determined is a left front door, a left front fender, a left front window, and a left front light, it is determined that the shooting angle of the to-be-determined photo is a first preset angle;
- the shooting angle of the to-be-determined photo is determined to be a second preset angle
- a part of the vehicle to be analyzed is a right front door, a right front fender, a right front window, and a right front light, it is determined that the shooting angle of the to-be-determined photo is a third preset angle;
- the shooting angle of the to-be-determined photo is determined to be a fourth preset angle.
- the technical solution of the present application which is essential or contributes to the prior art, may be embodied in the form of a software product stored in a storage medium (such as ROM/RAM as described above). , a disk, an optical disk, including a number of instructions for causing a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) to perform the methods described in the various embodiments of the present application.
- a terminal device which may be a mobile phone, a computer, a server, or a network device, etc.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
A vehicle damage determination method, a server, and a storage medium. The method comprises receiving a damage determination request and photos to be subjected to damage determination uploaded by a user (S10); then analyzing the photos to be subjected to damage determination by using a pre-trained classification model to obtain classification information of each part of a vehicle in each photo to be subjected to damage determination (S20); then determining a photographing angle of each photo to be subjected to damage determination according to the classification information of each part of the vehicle in each photo to be subjected to damage determination by using a pre-determined photographing angle determination rule (S30); and finally, obtaining a damage condition of each part of the vehicle in each photo to be subjected to damage determination by means of analysis using a pre-trained damage determination model by combining the photographing angle and the classification information of the part of the vehicle, and outputting a damage analysis result of each photo to be subjected to damage determination (S40). The use of the present method can effectively reduce labor power and material resources in a vehicle insurance claim settlement stage and improve the vehicle damage investigation accuracy and recall rate.
Description
本申请要求于2018年04月26日提交中国专利局、申请号为201810382312.1,名称为“车损判定方法、服务器及存储介质”的中国专利申请的优先权,该中国专利申请的整体内容以参考的方式结合本申请中。This application claims priority to Chinese Patent Application No. 201810382312.1, entitled "Car Damage Determination Method, Server and Storage Media", filed on April 26, 2018, the entire contents of which are incorporated by reference. The way it is combined with this application.
本申请涉及图片识别技术领域,尤其涉及一种车损判定方法、服务器及计算机可读存储介质。The present application relates to the field of picture recognition technologies, and in particular, to a vehicle loss determination method, a server, and a computer readable storage medium.
随着人工智能(Artificial Intelligence,AI)的发展技术的不断成熟,其应用领域也不断扩大。全球知名的科技公司都在各个领域中做出了关于人工智能的产业布局。在保险领域,保险公司也抓准此次机遇,利用人工智能解决现有保险业务的难点,实现产业的转型升级。对于保险业务而言,车险业务是其中一个重要的组成部分,而目前车险业务的一个难点在于车险理赔环节需要投入大量的人力、物力进行车损勘查。为了有效降低车险理赔环节的人力、物力。目前部分保险公司接受用户利用手持终端在事故现场拍摄车损照片,上传至车险理赔服务器进行自动检测,发起自助车险理赔。然而,目前现有的自动检测方案中,用户拍摄的车损照片经常因拍摄角度问题导致容易受损部位检测遗漏甚至识别错误,查全率及识别的准确率低。With the continuous development of artificial intelligence (AI) technology, its application fields are also expanding. The world's leading technology companies have made industrial layouts on artificial intelligence in various fields. In the insurance field, insurance companies have also seized this opportunity to use artificial intelligence to solve the difficulties of the existing insurance business and achieve industrial transformation and upgrading. For the insurance business, the auto insurance business is an important part of it. At present, one of the difficulties in the auto insurance business is that the auto insurance claims link needs to invest a lot of manpower and material resources for vehicle damage survey. In order to effectively reduce the manpower and material resources of the car insurance claims. At present, some insurance companies accept the use of handheld terminals to take pictures of car damage at the scene of the accident, upload them to the auto insurance claims server for automatic detection, and initiate self-service insurance claims. However, in the current automatic detection scheme, the photograph of the vehicle damage photographed by the user often causes omission or even identification error of the damaged portion due to the shooting angle problem, and the recall rate and the recognition accuracy are low.
发明内容Summary of the invention
鉴于以上内容,本申请提供一种车损判定方法、服务器及计算机可读存储介质,其主要目的在于提高车辆受损部位检测的全面性及准确性。In view of the above, the present application provides a vehicle damage determination method, a server, and a computer readable storage medium, the main purpose of which is to improve the comprehensiveness and accuracy of detection of damaged parts of a vehicle.
为实现上述目的,本申请提供一种车损判定方法,该方法包括:To achieve the above objective, the present application provides a vehicle damage determination method, including:
接收步骤:接收定损请求及用户上传的待定损照片;Receiving step: receiving a loss request and a photo to be determined by the user to be determined;
分类步骤:利用预先训练的分类模型对所述待定损照片进行分析,获取每张待定损照片中车辆各个部位的分类信息;Classification step: analyzing the to-be-determined loss photo by using a pre-trained classification model, and obtaining classification information of each part of the vehicle in each photo to be determined;
角度判定步骤:根据每张待定损照片中车辆各个部位的分类信息,利用 预先确定的拍摄角度判定规则,判断每张待定损照片的拍摄角度;Angle determining step: determining, according to the classification information of each part of the vehicle in each photo to be determined, using a predetermined shooting angle determination rule to determine the shooting angle of each photo to be determined;
定损步骤:结合拍摄角度以及车辆部位的分类信息,利用预先训练的定损模型分析出每张待定损照片中车辆各个部位的损伤情况,输出每张待定损照片中车辆的损伤分析结果。Determining the damage: Combining the shooting angle and the classification information of the vehicle part, using the pre-trained fixed loss model to analyze the damage of each part of the vehicle in each photo to be determined, and output the damage analysis result of the vehicle in each photo to be determined.
此外,本申请还提供一种服务器,该服务器包括:存储器及处理器,所述存储器上存储车损判定程序,所述车损判定程序被所述处理器执行,可实现如下步骤:In addition, the present application further provides a server, the server includes: a memory and a processor, wherein the memory stores a vehicle damage determination program, where the vehicle damage determination program is executed by the processor, and the following steps can be implemented:
接收步骤:接收定损请求及用户上传的待定损照片;Receiving step: receiving a loss request and a photo to be determined by the user to be determined;
分类步骤:利用预先训练的分类模型对所述待定损照片进行分析,获取每张待定损照片中车辆各个部位的分类信息;Classification step: analyzing the to-be-determined loss photo by using a pre-trained classification model, and obtaining classification information of each part of the vehicle in each photo to be determined;
角度判定步骤:根据每张待定损照片中车辆各个部位的分类信息,利用预先确定的拍摄角度判定规则,判断每张待定损照片的拍摄角度;The angle determining step: determining, according to the classification information of each part of the vehicle in each photo to be determined, the shooting angle of each photo to be determined by using a predetermined shooting angle determination rule;
定损步骤:结合拍摄角度以及车辆部位的分类信息,利用预先训练的定损模型分析出每张待定损照片中车辆各个部位的损伤情况,输出每张待定损照片中车辆的损伤分析结果。Determining the damage: Combining the shooting angle and the classification information of the vehicle part, using the pre-trained fixed loss model to analyze the damage of each part of the vehicle in each photo to be determined, and output the damage analysis result of the vehicle in each photo to be determined.
此外,为实现上述目的,本申请还提供一种计算机可读存储介质,所述计算机可读存储介质中包括车损判定程序,所述车损判定程序被处理器执行时,可实现如上所述车损判定方法中的任意步骤。In addition, in order to achieve the above object, the present application further provides a computer readable storage medium including a vehicle damage determination program, where the vehicle damage determination program is executed by a processor, as described above Any step in the vehicle damage determination method.
本申请提出的车损判定方法、服务器及计算机可读存储介质,通过利用分类模型对用户上传的待定损照片中车辆各个部位进行分类,接着利用预先确定的拍摄角度判定规则判断该照片的拍摄角度,最后车辆部位分类信息和拍摄角度,利用定损模型分析对车辆的部位进行损伤分析,反馈损伤分析结果,从而全面的对车辆进行检测,提高检测精准度。The vehicle damage determination method, the server and the computer readable storage medium proposed by the present application classify each part of the vehicle in the to-be-determined loss picture uploaded by the user by using the classification model, and then determine the shooting angle of the photo by using a predetermined shooting angle determination rule. Finally, the classification information and shooting angle of the vehicle parts are analyzed by the damage model to analyze the damage of the vehicle parts, and the damage analysis results are fed back to comprehensively test the vehicle to improve the detection accuracy.
图1为本申请服务器较佳实施例的示意图;1 is a schematic diagram of a preferred embodiment of a server of the present application;
图2为图1中车损判定程序较佳实施例的模块示意图;2 is a block diagram showing a preferred embodiment of the vehicle damage determination program of FIG. 1;
图3为本申请车损判定方法较佳实施例的流程图;3 is a flow chart of a preferred embodiment of a method for determining a vehicle damage according to the present application;
图4为本申请分类模型训练的流程图;4 is a flow chart of training of a classification model of the present application;
图5为本申请定损模型训练的流程图。FIG. 5 is a flow chart of the training of the fixed loss model of the present application.
本申请目的的实现、功能特点及优点将结合实施例,参照附图做进一步 说明。The implementation, functional features, and advantages of the present application will be further described with reference to the accompanying drawings.
应当理解,此处所描述的具体实施例仅用以解释本申请,并不用于限定本申请。It is understood that the specific embodiments described herein are merely illustrative of the application and are not intended to be limiting.
如图1所示,是本申请服务器1较佳实施例的示意图。As shown in FIG. 1, it is a schematic diagram of a preferred embodiment of the server 1 of the present application.
在本实施例中,服务器1是指车险理赔服务器,该服务器1可以是服务器、智能手机、平板电脑、个人电脑、便携计算机以及其它具有运算功能的电子设备。In the present embodiment, the server 1 refers to a car insurance claim server, which may be a server, a smart phone, a tablet computer, a personal computer, a portable computer, and other electronic devices having computing functions.
该服务器1包括:存储器11、处理器12、网络接口13及通信总线14。其中,网络接口13可选地可以包括标准的有线接口、无线接口(如WI-FI接口)。通信总线14用于实现这些组件之间的连接通信。The server 1 includes a memory 11, a processor 12, a network interface 13, and a communication bus 14. The network interface 13 can optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). Communication bus 14 is used to implement connection communication between these components.
存储器11至少包括一种类型的可读存储介质。所述至少一种类型的可读存储介质可为如闪存、硬盘、多媒体卡、卡型存储器等的非易失性存储介质。在一些实施例中,所述存储器11可以是所述服务器1的内部存储单元,例如该服务器1的硬盘。在另一些实施例中,所述存储器11也可以是所述服务器1的外部存储单元,例如所述服务器1上配备的插接式硬盘,智能存储卡(Smart Media Card,SMC),安全数字(Secure Digital,SD)卡,闪存卡(Flash Card)等。The memory 11 includes at least one type of readable storage medium. The at least one type of readable storage medium may be a non-volatile storage medium such as a flash memory, a hard disk, a multimedia card, a card type memory, or the like. In some embodiments, the memory 11 may be an internal storage unit of the server 1, such as a hard disk of the server 1. In other embodiments, the memory 11 may also be an external storage unit of the server 1, such as a plug-in hard disk equipped on the server 1, a smart memory card (SMC), and a secure digital ( Secure Digital, SD) cards, flash cards, etc.
在本实施例中,所述存储器11不仅可以用于存储安装于所述服务器1的应用软件及各类数据,例如车损判定程序10等。In the present embodiment, the memory 11 can be used not only for storing application software and various types of data installed in the server 1, such as the vehicle damage determination program 10 and the like.
处理器12在一些实施例中可以是一中央处理器(Central Processing Unit,CPU),微处理器或其它数据处理芯片,用于运行存储器11中存储的程序代码或处理数据,例如执行车损判定程序10的计算机程序代码、分类模型及定损模型的训练等。The processor 12, in some embodiments, may be a Central Processing Unit (CPU), microprocessor or other data processing chip for running program code or processing data stored in the memory 11, such as performing vehicle damage determination. The computer program code of the program 10, the classification model, and the training of the fixed loss model.
图1仅示出了具有组件11-14以及车损判定程序10的服务器1,但是应理解的是,并不要求实施所有示出的组件,可以替代的实施更多或者更少的组件。Figure 1 shows only server 1 with components 11-14 and vehicle damage determination program 10, but it should be understood that not all illustrated components may be implemented and that more or fewer components may be implemented instead.
可选地,该服务器1还可以包括显示器,显示器可以称为显示屏或显示单元。在一些实施例中显示器可以是LED显示器、液晶显示器、触控式液晶 显示器以及有机发光二极管(Organic Light-Emitting Diode,OLED)触摸器等。显示器用于显示在服务器1中处理的信息以及用于显示可视化的工作界面,例如显示车辆各个部位的损伤程度。Optionally, the server 1 may also include a display, which may be referred to as a display screen or a display unit. In some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an Organic Light-Emitting Diode (OLED) touch sensor. The display is used to display information processed in the server 1 and a work interface for displaying visualizations, such as showing the extent of damage to various parts of the vehicle.
可选地,该服务器1还可以包括用户接口,用户接口可以包括输入单元比如键盘(Keyboard)、语音输出装置比如音响、耳机等,可选地用户接口还可以包括标准的有线接口、无线接口。Optionally, the server 1 may further include a user interface, and the user interface may include an input unit such as a keyboard, a voice output device such as an audio, a headset, etc., optionally, the user interface may further include a standard wired interface and a wireless interface.
该服务器1还可以包括射频(Radio Frequency,RF)电路、传感器和音频电路等等,在此不再赘述。The server 1 may also include radio frequency (RF) circuits, sensors, audio circuits, and the like, and details are not described herein.
如图2所示,是图1中车损判定程序10较佳实施例的模块示意图。本申请所称的模块是指能够完成特定功能的一系列计算机程序指令段。2 is a block diagram of a preferred embodiment of the vehicle damage determination program 10 of FIG. A module as referred to in this application refers to a series of computer program instructions that are capable of performing a particular function.
在本实施例中,车损判定程序10包括:接收模块110、分类模块120、角度判定模块130、定损模块140,所述模块110-140所实现的功能或操作步骤均与上文类似,此处不再详述,示例性地,例如其中:In this embodiment, the vehicle damage determination program 10 includes: a receiving module 110, a classification module 120, an angle determining module 130, and a loss detecting module 140. The functions or operating steps implemented by the modules 110-140 are similar to the above. No more details here, exemplarily, for example:
接收模块110,用于接收定损请求及用户上传的待定损照片。其中,所述定损请求及车辆的待定损照片是用户通过手持终端向服务器1发出的。所述待定损照片是指待定损车辆全景的照片。例如,用户用手机的车险理赔应用程序APP上传待定损车辆全景的照片,并发起自助车险理赔。The receiving module 110 is configured to receive the loss request and the to-be-determined loss picture uploaded by the user. The fixed loss request and the to-be-determined loss photo of the vehicle are sent by the user to the server 1 through the handheld terminal. The photo to be determined refers to a photo of the panorama of the vehicle to be damaged. For example, the user uploads a photo of the panoramic view of the vehicle to be damaged using the mobile insurance claim application APP of the mobile phone, and initiates a self-service insurance claim.
分类模块120,用于利用预先训练的分类模型对所述待定损照片进行分析,获取每张待定损照片中车辆各个部位的分类信息。其中,所述车辆各个部位的分类信息包括:左前车门、左前叶子板、左前车窗、左前车灯、左后车门、左后叶子板、左后车窗、左后车灯、右前车门、右前叶子板、右前车窗、右前车灯、右后车门、右后叶子板、右后车窗、右后车灯、前车窗、后车窗、前车牌、后车牌及车辆识别码。The classification module 120 is configured to analyze the to-be-determined photo by using a pre-trained classification model, and obtain classification information of each part of the vehicle in each photo to be determined. The classification information of each part of the vehicle includes: a left front door, a left front fender, a left front window, a left front light, a left rear door, a left rear fender, a left rear window, a left rear light, a right front door, a right front Leaf panel, right front window, right front light, right rear door, right rear fender, right rear window, right rear light, front window, rear window, front license plate, rear license plate and vehicle identification code.
所述预先训练的分类模型为SSD模型,所述分类模型的具体结构如表1所示:The pre-trained classification model is an SSD model, and the specific structure of the classification model is as shown in Table 1:
表1:分类模型的网络结构Table 1: Network structure of the classification model
其中,Layer Name列表示每一层的名称,Input表示输入层,Conv表示模型的卷积层,Conv1表示模型的第1个卷积层,MaxPool表示模型的最大值池化层,MaxPool1表示模型的第1个最大值池化层,Fc表示模型中的全连接层,Fc1表示模型中第1个全连接层,Softmax表示Softmax分类器;Batch Size表示当前层的输入图像数目;Kernel Size表示当前层卷积核的尺度(例如,Kernel Size可以等于3,表示卷积核的尺度为3*3);Stride Size表示卷积核的移动步长,即做完一次卷积之后移动到下一个卷积位置的距离;Pad Size表示 对当前网络层之中的图像填充的大小;norm代表将feature map上的点归一化后得到的层;mbox_loc用来预测bounding box的回归值;mbox_conf用来预测feature map上每一个点的类别;mbox_priorbox用来生成bounding box;mbox_loss用来计算bounding box的损失函数。Among them, the Layer Name column indicates the name of each layer, Input indicates the input layer, Conv indicates the convolution layer of the model, Conv1 indicates the first convolution layer of the model, MaxPool indicates the maximum pooling layer of the model, and MaxPool1 indicates the model. The first maximum pooling layer, Fc represents the fully connected layer in the model, Fc1 represents the first fully connected layer in the model, Softmax represents the Softmax classifier; Batch Size represents the number of input images of the current layer; Kernel Size represents the current layer The scale of the convolution kernel (for example, Kernel Size can be equal to 3, indicating that the scale of the convolution kernel is 3*3); Stride Size indicates the moving step size of the convolution kernel, that is, moving to the next convolution after completing one convolution The distance of the position; Pad Size indicates the size of the image fill in the current network layer; norm represents the layer obtained by normalizing the points on the feature map; mbox_loc is used to predict the regression value of the bounding box; mbox_conf is used to predict the feature The category of each point on the map; mbox_priorbox is used to generate the bounding box; mbox_loss is used to calculate the loss function of the bounding box.
角度判定模块130,用于根据每张待定损照片中车辆各个部位的分类信息,利用预先确定的拍摄角度判定规则,判断每张待定损照片的拍摄角度。其中,所述预先确定的拍摄角度判定规则包括:The angle determining module 130 is configured to determine, according to the classification information of each part of the vehicle in each photo to be determined, the shooting angle of each photo to be determined by using a predetermined shooting angle determination rule. The predetermined shooting angle determination rule includes:
若一张待定损照片被分析出的车辆部位包括左前车门、左前叶子板、左前车窗、左前车灯,则判断该待定损照片的拍摄角度为第一预设角度;If a part of the vehicle to be determined to be determined is a left front door, a left front fender, a left front window, and a left front light, it is determined that the shooting angle of the to-be-determined photo is a first preset angle;
若一张待定损照片被分析出的车辆部位包括左后车门、左后叶子板、左后车窗、左后车灯,则判断该待定损照片的拍摄角度为第二预设角度;If a part of the vehicle to be analyzed is a left rear door, a left rear fender, a left rear window, and a left rear light, the shooting angle of the to-be-determined photo is determined to be a second preset angle;
若一张待定损照片被分析出的车辆部位包括右前车门、右前叶子板、右前车窗、右前车灯,则判断该待定损照片的拍摄角度为第三预设角度;If a part of the vehicle to be analyzed is a right front door, a right front fender, a right front window, and a right front light, it is determined that the shooting angle of the to-be-determined photo is a third preset angle;
若一张待定损照片被分析出的车辆部位包括右后车门、右后叶子板、右后车窗、右后车灯,则判断该待定损照片的拍摄角度为第四预设角度。If a part of the vehicle to be analyzed is a right rear door, a right rear fender, a right rear window, and a right rear light, the shooting angle of the to-be-determined photo is determined to be a fourth preset angle.
所述预设角度包括第一预设角度、第二预设角度、第三预设角度及第四预设角度。例如,预设角度包括左前45度、左后45度、右前45度及右后45度。进一步地,若拍摄角度中分析出未包含某个预设角度,则服务器1向用户手持终端发送提示信息:缺少**45度的待定损照片,请及时上传。The preset angle includes a first preset angle, a second preset angle, a third preset angle, and a fourth preset angle. For example, the preset angle includes 45 degrees to the front, 45 degrees to the left, 45 degrees to the front, and 45 degrees to the right. Further, if it is analyzed that the shooting angle does not include a certain preset angle, the server 1 sends a prompt message to the user handheld terminal: if the **45 degree pending loss photo is missing, please upload it in time.
定损模块140,用于结合拍摄角度以及车辆部位的分类信息,利用预先训练的定损模型分析出车辆各个部位的损伤情况,输出每张待定损照片中车辆的损伤分析结果。根据拍摄角度和车辆部位的分类信息,针对每个车辆部位,利用对应的定损模型进行损伤情况分析。例如,针对左前车门的部位,使用左前车门的定损模型进行训练。所述对应的定损模型是预先训练好的,该定损模型为VGG-16模型,所述定损模型的具体结构如表2所示:The damage module 140 is configured to analyze the damage situation of each part of the vehicle by using the pre-trained fixed loss model in combination with the shooting angle and the classification information of the vehicle part, and output the damage analysis result of the vehicle in each photo to be determined. According to the shooting angle and the classification information of the vehicle part, the damage situation analysis is performed for each vehicle part by using the corresponding fixed loss model. For example, for the part of the left front door, the training is performed using the fixed loss model of the left front door. The corresponding fixed loss model is pre-trained, and the fixed loss model is a VGG-16 model, and the specific structure of the fixed loss model is as shown in Table 2:
表2:定损模型的网络结构Table 2: Network structure of the fixed loss model
其中:Layer Name列表示每一层的名称,Input表示输入层,Conv表示模型的卷积层,Conv1表示模型的第1个卷积层,MaxPool表示模型的最大值池化层,MaxPool1表示模型的第1个最大值池化层,Fc表示模型中的全连接层,Fc1表示模型中第1个全连接层,Softmax表示Softmax分类器;Batch Size表示当前层的输入图像数目;Kernel Size表示当前层卷积核的尺度(例如,Kernel Size可以等于3,表示卷积核的尺度为3*3);Stride Size表示卷积核的移动步长,即做完一次卷积之后移动到下一个卷积位置的距离;Pad Size表示对当前网络层之中的图像填充的大小。Among them: Layer Name column indicates the name of each layer, Input indicates the input layer, Conv indicates the convolution layer of the model, Conv1 indicates the first convolution layer of the model, MaxPool indicates the maximum pooling layer of the model, and MaxPool1 indicates the model. The first maximum pooling layer, Fc represents the fully connected layer in the model, Fc1 represents the first fully connected layer in the model, Softmax represents the Softmax classifier; Batch Size represents the number of input images of the current layer; Kernel Size represents the current layer The scale of the convolution kernel (for example, Kernel Size can be equal to 3, indicating that the scale of the convolution kernel is 3*3); Stride Size indicates the moving step size of the convolution kernel, that is, moving to the next convolution after completing one convolution The distance of the location; Pad Size indicates the size of the image fill in the current network layer.
如图3所示,是本申请车损判定方法较佳实施例的流程图。As shown in FIG. 3, it is a flowchart of a preferred embodiment of the vehicle damage determination method of the present application.
在实施方式中,以自助车险理赔为例阐述本申请提供的车损判定方法的技术构思,其他类型的业务同样适用。In the embodiment, the technical concept of the vehicle damage determination method provided by the present application is explained by taking the self-service insurance claim as an example, and other types of services are also applicable.
在本实施例中,处理器12执行存储器11中存储的车损判定程序10的计算机程序时实现车损判定方法包括:步骤S10-步骤S40:In the present embodiment, the method for implementing the vehicle damage determination when the processor 12 executes the computer program of the vehicle damage determination program 10 stored in the memory 11 includes: Step S10 - Step S40:
步骤S10,接收模块110接收用户发送的定损请求及上传的车辆的待定损照片。当用户的车辆发生事故时,用户可以利用手机在事故现场拍摄车辆全景的车损照片,并将照片上传至车险理赔应用程序APP,发起自助车险理赔。所述车损照片从预设角度对车辆全景进行拍摄,防止遗漏检测受损部位。所述预设角度包括第一预设角度、第二预设角度、第三预设角度及第四预设角 度。例如,第一预设角度、第二预设角度、第三预设角度及第四预设角度分别对应的是左前45度、左后45度、右前45度及右后45度。In step S10, the receiving module 110 receives the fixed loss request sent by the user and the to-be-determined loss picture of the uploaded vehicle. When the user's vehicle has an accident, the user can use the mobile phone to take a picture of the vehicle's panoramic car damage at the scene of the accident, and upload the photo to the auto insurance claim application APP to initiate the self-service insurance claim. The vehicle damage photo captures the panoramic view of the vehicle from a preset angle to prevent missing detection of the damaged portion. The preset angle includes a first preset angle, a second preset angle, a third preset angle, and a fourth preset angle. For example, the first preset angle, the second preset angle, the third preset angle, and the fourth preset angle respectively correspond to a left front 45 degrees, a left rear 45 degrees, a right front 45 degrees, and a right rear 45 degrees.
步骤S20,分类模块120利用预先训练的分类模型对用户上传的待定损照片进行分析,获取每张待定损照片中车辆各个部位的分类信息。其中,所述分类模型为SSD模型,所述分类模型是预先训练好的。如图4所示,是本申请分类模型训练的流程图,该模型的训练步骤如下:In step S20, the classification module 120 analyzes the to-be-determined loss photos uploaded by the user by using the pre-trained classification model, and acquires classification information of each part of the vehicle in each photo to be determined. Wherein, the classification model is an SSD model, and the classification model is pre-trained. As shown in FIG. 4, it is a flowchart of the training of the classification model of the present application, and the training steps of the model are as follows:
获取第一预设数量,如10万张包含车辆的样本图片,并在每张样本图片上对车辆的车牌区域、车辆识别码区域等部位区域进行分类标注。所述分类标注是指使用不同颜色的框线分别框出不同的车辆部位,并对各个框线区域进行分类标注。其中,每种颜色的框线对应一个部位区域。例如,绿色的框线对应前车牌区域,红色的框线对应车辆识别码区域。Obtaining a first preset quantity, for example, 100,000 sample images including the vehicle, and classifying the vehicle license plate area, the vehicle identification code area, and the like on each sample picture. The classification labeling means that different vehicle parts are respectively framed by using frame lines of different colors, and each frame line area is classified and labeled. Wherein, the frame line of each color corresponds to a part area. For example, the green frame line corresponds to the front license plate area, and the red frame line corresponds to the vehicle identification code area.
将分类标注后的样本图片分成第一比例的训练集和第二比例的验证集,其中,第一比例大于第二比例。例如,随机将80%的样本图片,即8万张标注后的样本图片作为训练集,将剩余20%的样本图片,即2万张标注后的样本图片作为验证集。The classified sample picture is divided into a training set of a first ratio and a verification set of a second ratio, wherein the first ratio is greater than the second ratio. For example, 80% of the sample pictures, that is, 80,000 labeled sample pictures are randomly used as the training set, and the remaining 20% of the sample pictures, that is, 20,000 labeled sample pictures, are used as the verification set.
利用训练集中的样本图片对所述分类模型进行训练,并在训练完后利用验证集中的样本图片验证所述分类模型的准确率。例如,将训练集中8万张样本图片输入到SSD模型中训练,生成分类模型,并将验证集中2万张样本图片输入到生成的分类模型中验证该模型的准确率。The classification model is trained by using the sample picture in the training set, and the accuracy of the classification model is verified by using the sample picture in the verification set after the training. For example, 80,000 sample images in the training set are input into the SSD model for training, a classification model is generated, and 20,000 sample images in the verification set are input into the generated classification model to verify the accuracy of the model.
若准确率大于预设值,则训练完成,若准确率小于或等于预设值,则增加样本图片的数量,之后返回将样本图片分成训练集和验证集的步骤。假设,预设值为96%,若验证准确率大于96%,则训练完成,若准确率小于或等于96%,则增加2万张分类标注后的样本图片,之后返回将样本图片分成训练集和验证集的步骤。If the accuracy is greater than the preset value, the training is completed. If the accuracy is less than or equal to the preset value, the number of sample pictures is increased, and then the steps of dividing the sample picture into the training set and the verification set are returned. Assume that the default value is 96%. If the verification accuracy is greater than 96%, the training is completed. If the accuracy is less than or equal to 96%, add 20,000 sample images after classification, and then return to divide the sample image into training sets. And the steps of the validation set.
步骤S30,角度判定模块130根据每张待定损照片中车辆各个部位的分类信息,利用预先确定的拍摄角度判定规则,判断每张待定损照片的拍摄角度。所述预先确定的拍摄角度判定规则包括:In step S30, the angle determining module 130 determines the shooting angle of each photo to be determined by using the predetermined shooting angle determination rule according to the classification information of each part of the vehicle in the photo to be determined. The predetermined shooting angle determination rule includes:
若一张待定损照片被分析出的车辆部位包括左前车门、左前叶子板、左前车窗、左前车灯,则判断该待定损照片的拍摄角度为第一预设角度。例如,某张照片中包含左前车门、左前叶子板、左前车窗、左前车灯,则判断该照 片的拍摄角度为左前45度。If a part of the vehicle to be analyzed is included in the left front door, the left front fender, the left front window, and the left front light, it is determined that the shooting angle of the to-be-determined photo is the first preset angle. For example, if a photo contains a left front door, a left front fender, a left front window, and a left front light, it is judged that the photographing angle of the photograph is 45 degrees to the left.
若一张待定损照片被分析出的车辆部位包括左后车门、左后叶子板、左后车窗、左后车灯,则判断该待定损照片的拍摄角度为第二预设角度。例如,某张照片中包含左后车门、左后叶子板、左后车窗、左后车灯,则判断该照片的拍摄角度为左后45度。If a part of the vehicle to be analyzed is a left rear door, a left rear fender, a left rear window, and a left rear light, it is determined that the shooting angle of the to-be-determined photo is a second preset angle. For example, if a photo includes a left rear door, a left rear fender, a left rear window, and a left rear light, it is determined that the photographing angle of the photograph is 45 degrees to the left.
若一张待定损照片被分析出的车辆部位包括右前车门、右前叶子板、右前车窗、右前车灯,则判断该待定损照片的拍摄角度为第三预设角度。例如,某张照片中包含右前车门、右前叶子板、右前车窗、右前车灯,则判断该照片的拍摄角度为右前45度。If a part of the vehicle to be analyzed is a right front door, a right front fender, a right front window, and a right front light, it is determined that the shooting angle of the to-be-determined photo is a third preset angle. For example, if a photo includes a right front door, a right front fender, a right front window, and a right front light, it is determined that the photographing angle of the photograph is 45 degrees to the front.
若一张待定损照片被分析出的车辆部位包括右后车门、右后叶子板、右后车窗、右后车灯,则判断该待定损照片的拍摄角度为第四预设角度。例如,某张照片中包含右后车门、右后叶子板、右后车窗、右后车灯,则判断该照片的拍摄角度为右后45度。If a part of the vehicle to be analyzed is a right rear door, a right rear fender, a right rear window, and a right rear light, the shooting angle of the to-be-determined photo is determined to be a fourth preset angle. For example, if a photo includes a right rear door, a right rear fender, a right rear window, and a right rear light, it is determined that the photographing angle of the photograph is 45 degrees to the right.
进一步地,若角度判定模块130分析出拍摄角度中未包含某个预设角度,则服务器1向用户手持终端发送提示信息。例如,用户上传的照片中未包含右后车门、右后叶子板、右后车窗、右后车灯的照片,则角度判定模块130向用户手机发送提示信息:缺少右后45度的待定损照片,请及时上传。Further, if the angle determining module 130 analyzes that the shooting angle does not include a certain preset angle, the server 1 sends the prompt information to the user handheld terminal. For example, if the photo uploaded by the user does not include the photos of the right rear door, the right rear fender, the right rear window, and the right rear light, the angle determination module 130 sends a prompt message to the user's mobile phone: the missing right 45 degrees is to be determined. Please upload the photo in time.
步骤S40,定损模块140结合拍摄角度以及车辆部位的分类信息,利用预先训练的定损模型分析出每张待定损照片中车辆各个部位的损伤情况,输出每张待定损照片中车辆的损伤分析结果。所述结合拍摄角度以及车辆部位的分类信息是指针对车辆各个部位调用对应的定损模型对受损部位进行检测,结合拍摄角度判断受损部位的受损面积、受损程度以及赔偿金额。分析出车辆各个部位的受损情况后,以预设的方式提示用户。例如,通过短信的方式提示用户车辆各个部分的受损情况以及赔偿金额。其中,所述定损模型为VGG-16模型,所述定损模型是预先训练好的。如图5所示,是本申请定损模型训练的流程图,该模型的训练步骤如下:In step S40, the fixed loss module 140 combines the shooting angle and the classification information of the vehicle part, analyzes the damage situation of each part of the vehicle in each photo to be determined by using the pre-trained fixed loss model, and outputs the damage analysis of the vehicle in each photo to be determined. result. The combined shooting angle and the classification information of the vehicle part are used to detect the damaged part by calling the corresponding fixed loss model on each part of the vehicle, and determining the damaged area, the damage degree and the compensation amount of the damaged part according to the shooting angle. After analyzing the damage of various parts of the vehicle, the user is prompted in a preset manner. For example, the user is prompted by SMS to damage the various parts of the vehicle and the amount of compensation. Wherein, the fixed loss model is a VGG-16 model, and the fixed loss model is pre-trained. As shown in FIG. 5, it is a flowchart of the training of the fixed loss model of the present application, and the training steps of the model are as follows:
针对车辆的每个部位,获取第二预设数量,10万张该部位的样本图片,并对该部位的每张样本图片进行损伤标注,其中,该第二预设数量的样本图片包含第三比例的该部位存在损伤的样本图片和第四比例的该部位不存在损伤的样本图片。假设,第三比例为60%,第四比例为40%,则10万张该部位 的样本图片中包含6万张该部位存在损伤的样本图片和4万张该部位不存在损伤的样本图片。For each part of the vehicle, obtain a second preset number, 100,000 sample pictures of the part, and perform damage labeling on each sample picture of the part, wherein the second preset number of sample pictures includes the third The portion of the scale has a sample of the damaged sample and a sample of the sample at the fourth ratio where there is no damage. Assuming that the third ratio is 60% and the fourth ratio is 40%, 100,000 sample images of the part contain 60,000 samples of the damaged part of the part and 40,000 sample pictures of the part without damage.
在另一个实施例中,还可以设置不同程度的损伤等级,在对存在损伤的样本图片进行标注时还需要标注该样本图片的损伤等级,损伤等级包括严重损伤、较严重损伤、轻微损伤。同时在存在损伤的样本图片中为三个不同的损伤等级设置一定比例。例如,严重损伤占40%的存在损伤的样本图片,严重损伤占30%的存在损伤的样本图片,严重损伤占30%的存在损伤的样本图片。In another embodiment, different levels of damage may be set. When the image of the damaged sample is marked, the damage level of the sample image needs to be marked, and the damage level includes serious damage, serious damage, and slight damage. At the same time, a certain proportion is set for three different damage levels in the sample picture with damage. For example, severe damage accounts for 40% of the sample images with damage, serious damage accounts for 30% of the damaged sample images, and severe damage accounts for 30% of the damaged sample images.
将损伤标注后的该部位的样本图片随机分成第五比例的训练集和第六比例的验证集,其中,第五比例大于第六比例。例如,随机将70%损伤标注后的样本图片,即7万张标注后的样本图片作为训练集,将剩余30%损伤标注后的样本图片,即3万张标注后的样本图片作为验证集。The sample picture of the part after the damage is marked is randomly divided into a training set of a fifth ratio and a verification set of a sixth ratio, wherein the fifth ratio is greater than the sixth ratio. For example, the sample picture after 70% damage is randomly selected, that is, the 70,000-labeled sample picture is used as the training set, and the remaining 30% of the damaged sample picture, that is, the 30,000-labeled sample picture is used as the verification set.
利用训练集中的样本图片对所述定损模型进行训练,并在训练完后利用验证集中的样本图片验证所述定损模型的准确率。例如,将训练集中7万张损伤标注后的样本图片输入到SSD模型中训练,生成定损模型,并将验证集中2万张损伤标注后的样本图片输入到生成的定损模型中验证该模型的准确率。The fixed loss model is trained by using the sample picture in the training set, and the accuracy of the fixed loss model is verified by using the sample picture in the verification set after the training. For example, the sample images of the 70,000 damages in the training set are input into the SSD model for training, and a fixed loss model is generated, and the sample images of the 20,000 lesions in the verification set are input into the generated fixed loss model to verify the model. The accuracy rate.
若准确率大于预设阈值,则训练完成,若准确率小于或等于预设阈值,则增加该车辆部位的样本图片的数量,之后返回将样本图片分成训练集和验证集的步骤。假设,预设阈值为98%,若验证准确率大于98%,则训练完成,若准确率小于或等于98%,则增加2万张该部位损伤标注后的样本图片,之后返回将样本图片分成训练集和验证集的步骤。If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, the number of sample pictures of the vehicle part is increased, and then the step of dividing the sample picture into the training set and the verification set is returned. Assume that the preset threshold is 98%. If the verification accuracy is greater than 98%, the training is completed. If the accuracy is less than or equal to 98%, then 20,000 samples of the damaged label are added, and then the sample is divided into The steps of the training set and the validation set.
上述实施例提出的车损判定方法,通过接收上传的车辆全景的车损照片并利用分类模型对车损照片进行分类,识别出车辆各个部位,接着根据照片中的车辆部位信息判断拍摄角度,结合拍摄角度及车辆部位信息,利用对应的定损模型进行定损分析,输出损伤分析结果,降低分析的错误率,防止对受损部位检测遗漏。The vehicle damage determination method according to the above embodiment is configured to receive the vehicle damage photos of the uploaded vehicle panorama and classify the vehicle damage photos by using the classification model, identify each part of the vehicle, and then determine the shooting angle according to the vehicle part information in the photograph, and combine The shooting angle and the vehicle part information are used to perform the loss analysis using the corresponding fixed loss model, and the damage analysis result is output, and the analysis error rate is reduced to prevent the missing part from being detected.
此外,本申请实施例还提出一种计算机可读存储介质,所述计算机可读存储介质中包括车损判定程序10,所述车损判定程序10被处理器执行时实现如下操作:In addition, the embodiment of the present application further provides a computer readable storage medium, where the computer readable storage medium includes a vehicle damage determination program 10, and when the vehicle damage determination program 10 is executed by the processor, the following operations are implemented:
接收步骤:接收定损请求及用户上传的待定损照片;Receiving step: receiving a loss request and a photo to be determined by the user to be determined;
分类步骤:利用预先训练的分类模型对所述待定损照片进行分析,获取每张待定损照片中车辆各个部位的分类信息;Classification step: analyzing the to-be-determined loss photo by using a pre-trained classification model, and obtaining classification information of each part of the vehicle in each photo to be determined;
角度判定步骤:根据每张待定损照片中车辆各个部位的分类信息,利用预先确定的拍摄角度判定规则,判断每张待定损照片的拍摄角度;The angle determining step: determining, according to the classification information of each part of the vehicle in each photo to be determined, the shooting angle of each photo to be determined by using a predetermined shooting angle determination rule;
定损步骤:结合拍摄角度以及车辆部位的分类信息,利用预先训练的定损模型分析出每张待定损照片中车辆各个部位的损伤情况,输出每张待定损照片中车辆的损伤分析结果。Determining the damage: Combining the shooting angle and the classification information of the vehicle part, using the pre-trained fixed loss model to analyze the damage of each part of the vehicle in each photo to be determined, and output the damage analysis result of the vehicle in each photo to be determined.
优选地,所述分类模型为SSD模型,所述分类模型的训练步骤如下:Preferably, the classification model is an SSD model, and the training steps of the classification model are as follows:
获取第一预设数量的包含车辆的样本图片,并在每张样本图片上进行分类标注;Obtaining a first preset number of sample images containing the vehicle, and classifying each sample image;
将分类标注后的样本图片分成第一比例的训练集和第二比例的验证集,其中,第一比例大于第二比例;Dividing the classified sample picture into a training set of a first ratio and a verification set of a second ratio, wherein the first ratio is greater than the second ratio;
利用训练集中的样本图片对所述分类模型进行训练,并在训练完后利用验证集中的样本图片验证所述分类模型的准确率;The classification model is trained by using the sample picture in the training set, and the accuracy of the classification model is verified by using the sample picture in the verification set after the training;
若准确率大于预设值,则训练完成,若准确率小于或等于预设值,则增加样本图片的数量,之后返回将样本图片分成训练集和验证集的步骤。If the accuracy is greater than the preset value, the training is completed. If the accuracy is less than or equal to the preset value, the number of sample pictures is increased, and then the steps of dividing the sample picture into the training set and the verification set are returned.
优选地,所述定损模型为VGG-16模型,所述定损模型的训练步骤如下:Preferably, the fixed loss model is a VGG-16 model, and the training steps of the fixed loss model are as follows:
针对车辆的每个部位,获取第二预设数量的该部位的样本图片,并对该部位的每张样本图片进行损伤标注,其中,该第二预设数量的样本图片包含第三比例的该部位存在损伤的样本图片和第四比例的该部位不存在损伤的样本图片;Obtaining a second preset number of sample pictures of the part for each part of the vehicle, and performing damage labeling on each sample picture of the part, wherein the second preset number of sample pictures includes the third ratio of the a sample picture of the damaged part at the site and a sample picture of the fourth ratio where the part is not damaged;
将损伤标注后的该部位的样本图片随机分成第五比例的训练集和第六比例的验证集,其中,第五比例大于第六比例;The sample picture of the part after the damage label is randomly divided into a training set of a fifth ratio and a verification set of a sixth ratio, wherein the fifth ratio is greater than the sixth ratio;
利用训练集中的样本图片对所述定损模型进行训练,并在训练完后利用验证集中的样本图片验证所述定损模型的准确率;The determined loss model is trained by using the sample picture in the training set, and the accuracy of the fixed loss model is verified by using the sample picture in the verification set after the training;
若准确率大于预设阈值,则训练完成,若准确率小于或等于预设阈值,则增加该部位的样本图片的数量,之后返回将样本图片分成训练集和验证集的步骤。If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, the number of sample pictures of the part is increased, and then the step of dividing the sample picture into the training set and the verification set is returned.
优选地,该方法还包括:Preferably, the method further comprises:
提示步骤:若判断出的拍摄角度中未包含某个预设角度,则提醒用户重新拍摄并上传该预设角度的待定损照片。Prompt step: If the determined shooting angle does not include a preset angle, the user is reminded to re-shoot and upload the to-be-determined photo of the preset angle.
优选地,所述预设角度包括第一预设角度、第二预设角度、第三预设角度、第四预设角度,所述预先确定的拍摄角度判定规则包括:Preferably, the preset angle includes a first preset angle, a second preset angle, a third preset angle, and a fourth preset angle, and the predetermined shooting angle determination rule includes:
若一张待定损照片被分析出的车辆部位包括左前车门、左前叶子板、左前车窗、左前车灯,则判断该待定损照片的拍摄角度为第一预设角度;If a part of the vehicle to be determined to be determined is a left front door, a left front fender, a left front window, and a left front light, it is determined that the shooting angle of the to-be-determined photo is a first preset angle;
若一张待定损照片被分析出的车辆部位包括左后车门、左后叶子板、左后车窗、左后车灯,则判断该待定损照片的拍摄角度为第二预设角度;If a part of the vehicle to be analyzed is a left rear door, a left rear fender, a left rear window, and a left rear light, the shooting angle of the to-be-determined photo is determined to be a second preset angle;
若一张待定损照片被分析出的车辆部位包括右前车门、右前叶子板、右前车窗、右前车灯,则判断该待定损照片的拍摄角度为第三预设角度;If a part of the vehicle to be analyzed is a right front door, a right front fender, a right front window, and a right front light, it is determined that the shooting angle of the to-be-determined photo is a third preset angle;
若一张待定损照片被分析出的车辆部位包括右后车门、右后叶子板、右后车窗、右后车灯,则判断该待定损照片的拍摄角度为第四预设角度。If a part of the vehicle to be analyzed is a right rear door, a right rear fender, a right rear window, and a right rear light, the shooting angle of the to-be-determined photo is determined to be a fourth preset angle.
本申请之计算机可读存储介质的具体实施方式与上述车损判定方法的具体实施方式大致相同,在此不再赘述。The specific implementation manner of the computer readable storage medium of the present application is substantially the same as the specific implementation manner of the foregoing vehicle damage determination method, and details are not described herein again.
上述本申请实施例序号仅仅为了描述,不代表实施例的优劣。The serial numbers of the embodiments of the present application are merely for the description, and do not represent the advantages and disadvantages of the embodiments.
通过以上的实施方式的描述,本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现,当然也可以通过硬件,但很多情况下前者是更佳的实施方式。基于这样的理解,本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来,该计算机软件产品存储在如上所述的一个存储介质(如ROM/RAM、磁碟、光盘)中,包括若干指令用以使得一台终端设备(可以是手机,计算机,服务器,或者网络设备等)执行本申请各个实施例所述的方法。Through the description of the above embodiments, those skilled in the art can clearly understand that the foregoing embodiment method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be through hardware, but in many cases, the former is better. Implementation. Based on such understanding, the technical solution of the present application, which is essential or contributes to the prior art, may be embodied in the form of a software product stored in a storage medium (such as ROM/RAM as described above). , a disk, an optical disk, including a number of instructions for causing a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) to perform the methods described in the various embodiments of the present application.
以上仅为本申请的优选实施例,并非因此限制本申请的专利范围,凡是利用本申请说明书及附图内容所作的等效结构或等效流程变换,或直接或间接运用在其它相关的技术领域,均同理包括在本申请的专利保护范围内。The above is only a preferred embodiment of the present application, and thus does not limit the scope of the patent application, and the equivalent structure or equivalent process transformation made by the specification and the drawings of the present application, or directly or indirectly applied to other related technical fields. The same is included in the scope of patent protection of this application.
Claims (20)
- 一种车损判定方法,应用于服务器,其特征在于,所述方法包括:A vehicle damage determination method is applied to a server, wherein the method comprises:接收步骤:接收定损请求及用户上传的待定损照片;Receiving step: receiving a loss request and a photo to be determined by the user to be determined;分类步骤:利用预先训练的分类模型对所述待定损照片进行分析,获取每张待定损照片中车辆各个部位的分类信息;Classification step: analyzing the to-be-determined loss photo by using a pre-trained classification model, and obtaining classification information of each part of the vehicle in each photo to be determined;角度判定步骤:根据每张待定损照片中车辆各个部位的分类信息,利用预先确定的拍摄角度判定规则,判断每张待定损照片的拍摄角度;The angle determining step: determining, according to the classification information of each part of the vehicle in each photo to be determined, the shooting angle of each photo to be determined by using a predetermined shooting angle determination rule;定损步骤:结合拍摄角度以及车辆部位的分类信息,利用预先训练的定损模型分析出每张待定损照片中车辆各个部位的损伤情况,输出每张待定损照片中车辆的损伤分析结果。Determining the damage: Combining the shooting angle and the classification information of the vehicle part, using the pre-trained fixed loss model to analyze the damage of each part of the vehicle in each photo to be determined, and output the damage analysis result of the vehicle in each photo to be determined.
- 根据权利要求1所述的车损判定方法,其特征在于,所述分类模型为单次检测器模型,所述分类模型的训练步骤如下:The vehicle damage determination method according to claim 1, wherein the classification model is a single detector model, and the training steps of the classification model are as follows:获取第一预设数量的包含车辆的样本图片,并在每张样本图片上进行分类标注;Obtaining a first preset number of sample images containing the vehicle, and classifying each sample image;将分类标注后的样本图片分成第一比例的训练集和第二比例的验证集,其中,第一比例大于第二比例;Dividing the classified sample picture into a training set of a first ratio and a verification set of a second ratio, wherein the first ratio is greater than the second ratio;利用训练集中的样本图片对所述分类模型进行训练,并在训练完后利用验证集中的样本图片验证所述分类模型的准确率;The classification model is trained by using the sample picture in the training set, and the accuracy of the classification model is verified by using the sample picture in the verification set after the training;若准确率大于预设值,则训练完成,若准确率小于或等于预设值,则增加样本图片的数量,之后返回将样本图片分成训练集和验证集的步骤。If the accuracy is greater than the preset value, the training is completed. If the accuracy is less than or equal to the preset value, the number of sample pictures is increased, and then the steps of dividing the sample picture into the training set and the verification set are returned.
- 根据权利要求1所述的车损判定方法,其特征在于,所述定损模型为视觉几何组-16模型,所述定损模型的训练步骤如下:The vehicle damage determination method according to claim 1, wherein the fixed loss model is a visual geometric group-16 model, and the training steps of the fixed loss model are as follows:针对车辆的每个部位,获取第二预设数量的该部位的样本图片,并对该部位的每张样本图片进行损伤标注,其中,该第二预设数量的样本图片包含第三比例的该部位存在损伤的样本图片和第四比例的该部位不存在损伤的样本图片;Obtaining a second preset number of sample pictures of the part for each part of the vehicle, and performing damage labeling on each sample picture of the part, wherein the second preset number of sample pictures includes the third ratio of the a sample picture of the damaged part at the site and a sample picture of the fourth ratio where the part is not damaged;将损伤标注后的该部位的样本图片随机分成第五比例的训练集和第六比例的验证集,其中,第五比例大于第六比例;The sample picture of the part after the damage label is randomly divided into a training set of a fifth ratio and a verification set of a sixth ratio, wherein the fifth ratio is greater than the sixth ratio;利用训练集中的样本图片对所述定损模型进行训练,并在训练完后利用验证集中的样本图片验证所述定损模型的准确率;The determined loss model is trained by using the sample picture in the training set, and the accuracy of the fixed loss model is verified by using the sample picture in the verification set after the training;若准确率大于预设阈值,则训练完成,若准确率小于或等于预设阈值,则增加该部位的样本图片的数量,之后返回将样本图片分成训练集和验证集的步骤。If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, the number of sample pictures of the part is increased, and then the step of dividing the sample picture into the training set and the verification set is returned.
- 根据权利要求1或3所述的车损判定方法,其特征在于,所述定损步骤包括:The vehicle damage determination method according to claim 1 or 3, wherein the determining the loss comprises:根据照片中车辆的各个部位,调用对应的定损模型对受损部位进行检测,结合拍摄角度判断受损部位的受损面积、受损程度及赔偿金额。According to the various parts of the vehicle in the photo, the corresponding damage model is called to detect the damaged part, and the damaged area, the damage degree and the compensation amount are determined according to the shooting angle.
- 根据权利要求1所述的车损判定方法,其特征在于,该方法还包括:The vehicle damage determination method according to claim 1, wherein the method further comprises:提示步骤:若判断出的拍摄角度中未包含某个预设角度,则提醒用户重新拍摄并上传该预设角度的待定损照片。Prompt step: If the determined shooting angle does not include a preset angle, the user is reminded to re-shoot and upload the to-be-determined photo of the preset angle.
- 根据权利要求1或5所述的车损判定方法,其特征在于,所述预设角度包括第一预设角度、第二预设角度、第三预设角度、第四预设角度。The vehicle damage determination method according to claim 1 or 5, wherein the preset angle comprises a first preset angle, a second preset angle, a third preset angle, and a fourth preset angle.
- 根据权利要求1或5所述的车损判定方法,其特征在于,所述预先确定的拍摄角度判定规则包括:The vehicle damage determination method according to claim 1 or 5, wherein the predetermined shooting angle determination rule comprises:若一张待定损照片被分析出的车辆部位包括左前车门、左前叶子板、左前车窗、左前车灯,则判断该待定损照片的拍摄角度为第一预设角度;If a part of the vehicle to be determined to be determined is a left front door, a left front fender, a left front window, and a left front light, it is determined that the shooting angle of the to-be-determined photo is a first preset angle;若一张待定损照片被分析出的车辆部位包括左后车门、左后叶子板、左后车窗、左后车灯,则判断该待定损照片的拍摄角度为第二预设角度;If a part of the vehicle to be analyzed is a left rear door, a left rear fender, a left rear window, and a left rear light, the shooting angle of the to-be-determined photo is determined to be a second preset angle;若一张待定损照片被分析出的车辆部位包括右前车门、右前叶子板、右前车窗、右前车灯,则判断该待定损照片的拍摄角度为第三预设角度;If a part of the vehicle to be analyzed is a right front door, a right front fender, a right front window, and a right front light, it is determined that the shooting angle of the to-be-determined photo is a third preset angle;若一张待定损照片被分析出的车辆部位包括右后车门、右后叶子板、右后车窗、右后车灯,则判断该待定损照片的拍摄角度为第四预设角度。If a part of the vehicle to be analyzed is a right rear door, a right rear fender, a right rear window, and a right rear light, the shooting angle of the to-be-determined photo is determined to be a fourth preset angle.
- 一种服务器,其特征在于,所述服务器包括:存储器及处理器,所述存储器上存储有车损判定程序,所述车损判定程序被所述处理器执行,可实现如下步骤:A server, comprising: a memory and a processor, wherein the memory stores a vehicle damage determination program, wherein the vehicle damage determination program is executed by the processor, and the following steps can be implemented:接收步骤:接收定损请求及用户上传的待定损照片;Receiving step: receiving a loss request and a photo to be determined by the user to be determined;分类步骤:利用预先训练的分类模型对所述待定损照片进行分析,获取每张待定损照片中车辆各个部位的分类信息;Classification step: analyzing the to-be-determined loss photo by using a pre-trained classification model, and obtaining classification information of each part of the vehicle in each photo to be determined;角度判定步骤:根据每张待定损照片中车辆各个部位的分类信息,利用预先确定的拍摄角度判定规则,判断每张待定损照片的拍摄角度;The angle determining step: determining, according to the classification information of each part of the vehicle in each photo to be determined, the shooting angle of each photo to be determined by using a predetermined shooting angle determination rule;定损步骤:结合拍摄角度以及车辆部位的分类信息,利用预先训练的定损模型分析出每张待定损照片中车辆各个部位的损伤情况,输出每张待定损照片中车辆的损伤分析结果。Determining the damage: Combining the shooting angle and the classification information of the vehicle part, using the pre-trained fixed loss model to analyze the damage of each part of the vehicle in each photo to be determined, and output the damage analysis result of the vehicle in each photo to be determined.
- 根据权利要求8所述的服务器,其特征在于,所述分类模型为单次检测器模型,所述分类模型的训练步骤如下:The server according to claim 8, wherein the classification model is a single detector model, and the training steps of the classification model are as follows:获取第一预设数量的包含车辆的样本图片,并在每张样本图片上进行分类标注;Obtaining a first preset number of sample images containing the vehicle, and classifying each sample image;将分类标注后的样本图片分成第一比例的训练集和第二比例的验证集,其中,第一比例大于第二比例;Dividing the classified sample picture into a training set of a first ratio and a verification set of a second ratio, wherein the first ratio is greater than the second ratio;利用训练集中的样本图片对所述分类模型进行训练,并在训练完后利用验证集中的样本图片验证所述分类模型的准确率;The classification model is trained by using the sample picture in the training set, and the accuracy of the classification model is verified by using the sample picture in the verification set after the training;若准确率大于预设值,则训练完成,若准确率小于或等于预设值,则增加样本图片的数量,之后返回将样本图片分成训练集和验证集的步骤。If the accuracy is greater than the preset value, the training is completed. If the accuracy is less than or equal to the preset value, the number of sample pictures is increased, and then the steps of dividing the sample picture into the training set and the verification set are returned.
- 根据权利要求8所述的服务器,其特征在于,所述定损模型为视觉几何组-16模型,所述定损模型的训练步骤如下:The server according to claim 8, wherein the fixed loss model is a visual geometric group-16 model, and the training steps of the fixed loss model are as follows:针对车辆的每个部位,获取第二预设数量的该部位的样本图片,并对该部位的每张样本图片进行损伤标注,其中,该第二预设数量的样本图片包含第三比例的该部位存在损伤的样本图片和第四比例的该部位不存在损伤的样本图片;Obtaining a second preset number of sample pictures of the part for each part of the vehicle, and performing damage labeling on each sample picture of the part, wherein the second preset number of sample pictures includes the third ratio of the a sample picture of the damaged part at the site and a sample picture of the fourth ratio where the part is not damaged;将损伤标注后的该部位的样本图片随机分成第五比例的训练集和第六比例的验证集,其中,第五比例大于第六比例;The sample picture of the part after the damage label is randomly divided into a training set of a fifth ratio and a verification set of a sixth ratio, wherein the fifth ratio is greater than the sixth ratio;利用训练集中的样本图片对所述定损模型进行训练,并在训练完后利用验证集中的样本图片验证所述定损模型的准确率;The determined loss model is trained by using the sample picture in the training set, and the accuracy of the fixed loss model is verified by using the sample picture in the verification set after the training;若准确率大于预设阈值,则训练完成,若准确率小于或等于预设阈值,则增加该部位的样本图片的数量,之后返回将样本图片分成训练集和验证集的步骤。If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, the number of sample pictures of the part is increased, and then the step of dividing the sample picture into the training set and the verification set is returned.
- 根据权利要求8或10所述的服务器,其特征在于,所述定损步骤包括:The server according to claim 8 or 10, wherein the determining the loss comprises:根据照片中车辆的各个部位,调用对应的定损模型对受损部位进行检测,结合拍摄角度判断受损部位的受损面积、受损程度及赔偿金额。According to the various parts of the vehicle in the photo, the corresponding damage model is called to detect the damaged part, and the damaged area, the damage degree and the compensation amount are determined according to the shooting angle.
- 根据权利要求8所述的服务器,其特征在于,所述车损判定程序被 所述处理器执行,还可实现如下步骤:The server according to claim 8, wherein said vehicle damage determination program is executed by said processor, and the following steps are also implemented:提示步骤:若判断出的拍摄角度中未包含某个预设角度,则提醒用户重新拍摄并上传该预设角度的待定损照片。Prompt step: If the determined shooting angle does not include a preset angle, the user is reminded to re-shoot and upload the to-be-determined photo of the preset angle.
- 根据权利要求8或12所述的服务器,其特征在于,所述预设角度包括第一预设角度、第二预设角度、第三预设角度、第四预设角度。The server according to claim 8 or 12, wherein the preset angle comprises a first preset angle, a second preset angle, a third preset angle, and a fourth preset angle.
- 根据权利要求8或12所述的服务器,其特征在于,所述预先确定的拍摄角度判定规则包括:The server according to claim 8 or 12, wherein the predetermined shooting angle determination rule comprises:若一张待定损照片被分析出的车辆部位包括左前车门、左前叶子板、左前车窗、左前车灯,则判断该待定损照片的拍摄角度为第一预设角度;If a part of the vehicle to be determined to be determined is a left front door, a left front fender, a left front window, and a left front light, it is determined that the shooting angle of the to-be-determined photo is a first preset angle;若一张待定损照片被分析出的车辆部位包括左后车门、左后叶子板、左后车窗、左后车灯,则判断该待定损照片的拍摄角度为第二预设角度;If a part of the vehicle to be analyzed is a left rear door, a left rear fender, a left rear window, and a left rear light, the shooting angle of the to-be-determined photo is determined to be a second preset angle;若一张待定损照片被分析出的车辆部位包括右前车门、右前叶子板、右前车窗、右前车灯,则判断该待定损照片的拍摄角度为第三预设角度;If a part of the vehicle to be analyzed is a right front door, a right front fender, a right front window, and a right front light, it is determined that the shooting angle of the to-be-determined photo is a third preset angle;若一张待定损照片被分析出的车辆部位包括右后车门、右后叶子板、右后车窗、右后车灯,则判断该待定损照片的拍摄角度为第四预设角度。If a part of the vehicle to be analyzed is a right rear door, a right rear fender, a right rear window, and a right rear light, the shooting angle of the to-be-determined photo is determined to be a fourth preset angle.
- 一种计算机可读存储介质,其特征在于,所述计算机可读存储介质中包括车损判定程序,所述车损判定程序被处理器执行时,可实现如下步骤:A computer readable storage medium, comprising: a vehicle damage determination program, wherein when the vehicle damage determination program is executed by a processor, the following steps can be implemented:接收步骤:接收定损请求及用户上传的待定损照片;Receiving step: receiving a loss request and a photo to be determined by the user to be determined;分类步骤:利用预先训练的分类模型对所述待定损照片进行分析,获取每张待定损照片中车辆各个部位的分类信息;Classification step: analyzing the to-be-determined loss photo by using a pre-trained classification model, and obtaining classification information of each part of the vehicle in each photo to be determined;角度判定步骤:根据每张待定损照片中车辆各个部位的分类信息,利用预先确定的拍摄角度判定规则,判断每张待定损照片的拍摄角度;The angle determining step: determining, according to the classification information of each part of the vehicle in each photo to be determined, the shooting angle of each photo to be determined by using a predetermined shooting angle determination rule;定损步骤:结合拍摄角度以及车辆部位的分类信息,利用预先训练的定损模型分析出每张待定损照片中车辆各个部位的损伤情况,输出每张待定损照片中车辆的损伤分析结果。Determining the damage: Combining the shooting angle and the classification information of the vehicle part, using the pre-trained fixed loss model to analyze the damage of each part of the vehicle in each photo to be determined, and output the damage analysis result of the vehicle in each photo to be determined.
- 根据权利要求15所述的计算机可读存储介质,其特征在于,所述分类模型为单次检测器模型,所述分类模型的训练步骤如下:The computer readable storage medium according to claim 15, wherein the classification model is a single detector model, and the training steps of the classification model are as follows:获取第一预设数量的包含车辆的样本图片,并在每张样本图片上进行分类标注;Obtaining a first preset number of sample images containing the vehicle, and classifying each sample image;将分类标注后的样本图片分成第一比例的训练集和第二比例的验证集, 其中,第一比例大于第二比例;Dividing the classified sample picture into a training set of a first ratio and a verification set of a second ratio, wherein the first ratio is greater than the second ratio;利用训练集中的样本图片对所述分类模型进行训练,并在训练完后利用验证集中的样本图片验证所述分类模型的准确率;The classification model is trained by using the sample picture in the training set, and the accuracy of the classification model is verified by using the sample picture in the verification set after the training;若准确率大于预设值,则训练完成,若准确率小于或等于预设值,则增加样本图片的数量,之后返回将样本图片分成训练集和验证集的步骤。If the accuracy is greater than the preset value, the training is completed. If the accuracy is less than or equal to the preset value, the number of sample pictures is increased, and then the steps of dividing the sample picture into the training set and the verification set are returned.
- 根据权利要求15所述的计算机可读存储介质,其特征在于,所述定损模型为视觉几何组-16模型,所述定损模型的训练步骤如下:The computer readable storage medium according to claim 15, wherein the fixed loss model is a visual geometric group-16 model, and the training steps of the fixed loss model are as follows:针对车辆的每个部位,获取第二预设数量的该部位的样本图片,并对该部位的每张样本图片进行损伤标注,其中,该第二预设数量的样本图片包含第三比例的该部位存在损伤的样本图片和第四比例的该部位不存在损伤的样本图片;Obtaining a second preset number of sample pictures of the part for each part of the vehicle, and performing damage labeling on each sample picture of the part, wherein the second preset number of sample pictures includes the third ratio of the a sample picture of the damaged part at the site and a sample picture of the fourth ratio where the part is not damaged;将损伤标注后的该部位的样本图片随机分成第五比例的训练集和第六比例的验证集,其中,第五比例大于第六比例;The sample picture of the part after the damage label is randomly divided into a training set of a fifth ratio and a verification set of a sixth ratio, wherein the fifth ratio is greater than the sixth ratio;利用训练集中的样本图片对所述定损模型进行训练,并在训练完后利用验证集中的样本图片验证所述定损模型的准确率;The determined loss model is trained by using the sample picture in the training set, and the accuracy of the fixed loss model is verified by using the sample picture in the verification set after the training;若准确率大于预设阈值,则训练完成,若准确率小于或等于预设阈值,则增加该部位的样本图片的数量,之后返回将样本图片分成训练集和验证集的步骤。If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, the number of sample pictures of the part is increased, and then the step of dividing the sample picture into the training set and the verification set is returned.
- 根据权利要求15所述的计算机可读存储介质,其特征在于,所述车损判定程序被所述处理器执行,还可实现如下步骤:The computer readable storage medium according to claim 15, wherein said vehicle damage determination program is executed by said processor, and the following steps are further implemented:提示步骤:若判断出的拍摄角度中未包含某个预设角度,则提醒用户重新拍摄并上传该预设角度的待定损照片。Prompt step: If the determined shooting angle does not include a preset angle, the user is reminded to re-shoot and upload the to-be-determined photo of the preset angle.
- 根据权利要求15或18所述的计算机可读存储介质,其特征在于,所述预设角度包括第一预设角度、第二预设角度、第三预设角度、第四预设角度。The computer readable storage medium according to claim 15 or 18, wherein the preset angle comprises a first preset angle, a second preset angle, a third preset angle, and a fourth preset angle.
- 根据权利要求15或18所述的计算机可读存储介质,其特征在于,所述预先确定的拍摄角度判定规则包括:The computer readable storage medium according to claim 15 or 18, wherein the predetermined shooting angle determination rule comprises:若一张待定损照片被分析出的车辆部位包括左前车门、左前叶子板、左前车窗、左前车灯,则判断该待定损照片的拍摄角度为第一预设角度;If a part of the vehicle to be determined to be determined is a left front door, a left front fender, a left front window, and a left front light, it is determined that the shooting angle of the to-be-determined photo is a first preset angle;若一张待定损照片被分析出的车辆部位包括左后车门、左后叶子板、左 后车窗、左后车灯,则判断该待定损照片的拍摄角度为第二预设角度;If a part of the vehicle to be determined to be determined is a left rear door, a left rear fender, a left rear window, and a left rear light, it is determined that the shooting angle of the to-be-determined photo is a second preset angle;若一张待定损照片被分析出的车辆部位包括右前车门、右前叶子板、右前车窗、右前车灯,则判断该待定损照片的拍摄角度为第三预设角度;If a part of the vehicle to be analyzed is a right front door, a right front fender, a right front window, and a right front light, it is determined that the shooting angle of the to-be-determined photo is a third preset angle;若一张待定损照片被分析出的车辆部位包括右后车门、右后叶子板、右后车窗、右后车灯,则判断该待定损照片的拍摄角度为第四预设角度。If a part of the vehicle to be analyzed is a right rear door, a right rear fender, a right rear window, and a right rear light, the shooting angle of the to-be-determined photo is determined to be a fourth preset angle.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810382312.1A CN108734702A (en) | 2018-04-26 | 2018-04-26 | Vehicle damages determination method, server and storage medium |
CN2018103823121 | 2018-04-26 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2019205376A1 true WO2019205376A1 (en) | 2019-10-31 |
Family
ID=63939844
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2018/102121 WO2019205376A1 (en) | 2018-04-26 | 2018-08-24 | Vehicle damage determination method, server, and storage medium |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN108734702A (en) |
WO (1) | WO2019205376A1 (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111523615A (en) * | 2020-05-08 | 2020-08-11 | 深源恒际科技有限公司 | Assembly line closed loop flow method for realizing vehicle appearance professional damage labeling |
CN111553268A (en) * | 2020-04-27 | 2020-08-18 | 深圳壹账通智能科技有限公司 | Vehicle part identification method and device, computer equipment and storage medium |
CN111680746A (en) * | 2020-06-08 | 2020-09-18 | 平安科技(深圳)有限公司 | Vehicle loss detection model training method, vehicle loss detection device, vehicle loss detection equipment and vehicle loss detection medium |
CN112418214A (en) * | 2020-11-06 | 2021-02-26 | 北京五八信息技术有限公司 | Vehicle identification code identification method and device, electronic equipment and storage medium |
CN113160206A (en) * | 2021-04-30 | 2021-07-23 | 上海海事大学 | Method for detecting foreign matters on steel rail track |
CN113326954A (en) * | 2021-06-25 | 2021-08-31 | 中国平安财产保险股份有限公司 | Vehicle maintenance task scheduling method, device, equipment and storage medium |
US20210304153A1 (en) * | 2020-03-30 | 2021-09-30 | Lyft, Inc. | Utilizing a transportation matching system in conjunction with a multi-track vehicle service center to service transportation vehicles |
CN113640146A (en) * | 2021-07-21 | 2021-11-12 | 东风柳州汽车有限公司 | Vehicle canopy torsion resistance evaluation method, device, equipment and storage medium |
CN114140430A (en) * | 2021-11-30 | 2022-03-04 | 北京比特易湃信息技术有限公司 | Vehicle damage reporting method based on deep learning |
CN114898155A (en) * | 2022-05-18 | 2022-08-12 | 平安科技(深圳)有限公司 | Vehicle damage assessment method, device, equipment and storage medium |
Families Citing this family (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109344819A (en) * | 2018-12-13 | 2019-02-15 | 深源恒际科技有限公司 | Vehicle damage recognition methods based on deep learning |
CN110009509B (en) * | 2019-01-02 | 2021-02-19 | 创新先进技术有限公司 | Method and device for evaluating vehicle damage recognition model |
CN109919785A (en) * | 2019-02-22 | 2019-06-21 | 德联易控科技(北京)有限公司 | Assessment processing method, apparatus, equipment and the storage medium of car damage identification |
CN110443718B (en) * | 2019-07-17 | 2022-11-04 | 卓尔智联(武汉)研究院有限公司 | Vehicle insurance claim settlement method, computer device and readable storage medium |
CN110473418B (en) * | 2019-07-25 | 2022-05-20 | 平安科技(深圳)有限公司 | Dangerous road segment identification method, dangerous road segment identification device, server and storage medium |
CN110689440A (en) * | 2019-08-13 | 2020-01-14 | 平安科技(深圳)有限公司 | Vehicle insurance claim settlement identification method and device based on image identification, computer equipment and storage medium |
CN111191400B (en) * | 2019-12-31 | 2023-12-29 | 上海钧正网络科技有限公司 | Vehicle part life prediction method and system based on user fault reporting data |
CN111311540A (en) * | 2020-01-13 | 2020-06-19 | 平安科技(深圳)有限公司 | Vehicle damage assessment method and device, computer equipment and storage medium |
CN111259969A (en) * | 2020-01-19 | 2020-06-09 | 上海钧正网络科技有限公司 | Failure reporting identification method, device, server and medium |
CN111489433B (en) * | 2020-02-13 | 2023-04-25 | 北京百度网讯科技有限公司 | Method and device for positioning damage of vehicle, electronic equipment and readable storage medium |
CN111652087B (en) * | 2020-05-15 | 2023-07-18 | 泰康保险集团股份有限公司 | Car inspection method, device, electronic equipment and storage medium |
CN111666990A (en) * | 2020-05-27 | 2020-09-15 | 平安科技(深圳)有限公司 | Vehicle damage characteristic detection method and device, computer equipment and storage medium |
CN112465018B (en) * | 2020-11-26 | 2024-02-02 | 深源恒际科技有限公司 | Intelligent screenshot method and system of vehicle video damage assessment system based on deep learning |
CN113627252A (en) * | 2021-07-07 | 2021-11-09 | 浙江吉利控股集团有限公司 | Vehicle damage assessment method and device, storage medium and electronic equipment |
CN114241180A (en) * | 2021-12-15 | 2022-03-25 | 平安科技(深圳)有限公司 | Image detection method and device for vehicle damage claims, computer equipment and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100194562A1 (en) * | 2009-01-30 | 2010-08-05 | Jong-Moon Lee | Failure recognition system |
CN105719188A (en) * | 2016-01-22 | 2016-06-29 | 平安科技(深圳)有限公司 | Method and server for achieving insurance claim anti-fraud based on consistency of multiple pictures |
CN106780048A (en) * | 2016-11-28 | 2017-05-31 | 中国平安财产保险股份有限公司 | A kind of self-service Claims Resolution method of intelligent vehicle insurance, self-service Claims Resolution apparatus and system |
CN107092922A (en) * | 2017-03-13 | 2017-08-25 | 平安科技(深圳)有限公司 | Car damages recognition methods and server |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107194398B (en) * | 2017-05-10 | 2018-09-25 | 平安科技(深圳)有限公司 | Vehicle damages recognition methods and the system at position |
-
2018
- 2018-04-26 CN CN201810382312.1A patent/CN108734702A/en active Pending
- 2018-08-24 WO PCT/CN2018/102121 patent/WO2019205376A1/en active Application Filing
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100194562A1 (en) * | 2009-01-30 | 2010-08-05 | Jong-Moon Lee | Failure recognition system |
CN105719188A (en) * | 2016-01-22 | 2016-06-29 | 平安科技(深圳)有限公司 | Method and server for achieving insurance claim anti-fraud based on consistency of multiple pictures |
CN106780048A (en) * | 2016-11-28 | 2017-05-31 | 中国平安财产保险股份有限公司 | A kind of self-service Claims Resolution method of intelligent vehicle insurance, self-service Claims Resolution apparatus and system |
CN107092922A (en) * | 2017-03-13 | 2017-08-25 | 平安科技(深圳)有限公司 | Car damages recognition methods and server |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20210304153A1 (en) * | 2020-03-30 | 2021-09-30 | Lyft, Inc. | Utilizing a transportation matching system in conjunction with a multi-track vehicle service center to service transportation vehicles |
CN111553268A (en) * | 2020-04-27 | 2020-08-18 | 深圳壹账通智能科技有限公司 | Vehicle part identification method and device, computer equipment and storage medium |
CN111523615A (en) * | 2020-05-08 | 2020-08-11 | 深源恒际科技有限公司 | Assembly line closed loop flow method for realizing vehicle appearance professional damage labeling |
CN111523615B (en) * | 2020-05-08 | 2024-03-26 | 北京深智恒际科技有限公司 | Assembly line closed-loop flow method for realizing vehicle appearance professional damage labeling |
CN111680746A (en) * | 2020-06-08 | 2020-09-18 | 平安科技(深圳)有限公司 | Vehicle loss detection model training method, vehicle loss detection device, vehicle loss detection equipment and vehicle loss detection medium |
CN111680746B (en) * | 2020-06-08 | 2023-08-04 | 平安科技(深圳)有限公司 | Vehicle damage detection model training, vehicle damage detection method, device, equipment and medium |
CN112418214A (en) * | 2020-11-06 | 2021-02-26 | 北京五八信息技术有限公司 | Vehicle identification code identification method and device, electronic equipment and storage medium |
CN113160206B (en) * | 2021-04-30 | 2023-09-12 | 上海海事大学 | Method for detecting foreign matters on steel rail |
CN113160206A (en) * | 2021-04-30 | 2021-07-23 | 上海海事大学 | Method for detecting foreign matters on steel rail track |
CN113326954B (en) * | 2021-06-25 | 2023-07-07 | 中国平安财产保险股份有限公司 | Vehicle maintenance task scheduling method, device, equipment and storage medium |
CN113326954A (en) * | 2021-06-25 | 2021-08-31 | 中国平安财产保险股份有限公司 | Vehicle maintenance task scheduling method, device, equipment and storage medium |
CN113640146A (en) * | 2021-07-21 | 2021-11-12 | 东风柳州汽车有限公司 | Vehicle canopy torsion resistance evaluation method, device, equipment and storage medium |
CN114140430A (en) * | 2021-11-30 | 2022-03-04 | 北京比特易湃信息技术有限公司 | Vehicle damage reporting method based on deep learning |
CN114898155A (en) * | 2022-05-18 | 2022-08-12 | 平安科技(深圳)有限公司 | Vehicle damage assessment method, device, equipment and storage medium |
CN114898155B (en) * | 2022-05-18 | 2024-05-28 | 平安科技(深圳)有限公司 | Vehicle damage assessment method, device, equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN108734702A (en) | 2018-11-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2019205376A1 (en) | Vehicle damage determination method, server, and storage medium | |
US10984293B2 (en) | Image processing method and apparatus | |
US11127087B1 (en) | Insurance claim submission using captured images and optical character recognition (OCR) | |
WO2019205391A1 (en) | Apparatus and method for generating vehicle damage classification model, and computer readable storage medium | |
TWI709919B (en) | Auto insurance image processing method, device, server and system | |
CN107798299B (en) | Bill information identification method, electronic device and readable storage medium | |
WO2019174130A1 (en) | Bill recognition method, server, and computer readable storage medium | |
CN109815997B (en) | Method and related device for identifying vehicle damage based on deep learning | |
CN107194398B (en) | Vehicle damages recognition methods and the system at position | |
JP6250014B2 (en) | System and method for mobile image capture and processing | |
CN109886928B (en) | Target cell marking method, device, storage medium and terminal equipment | |
WO2020024407A1 (en) | Electronic device, method for intelligently processing car insurance claim, and storage medium | |
US10152639B2 (en) | Method and system for identifying a vacant parking space in a parking lot | |
WO2021135879A1 (en) | Vehicle data monitoring method and apparatus, computer device, and storage medium | |
US9349046B2 (en) | Smart optical input/output (I/O) extension for context-dependent workflows | |
CN108491866B (en) | Pornographic picture identification method, electronic device and readable storage medium | |
WO2020155790A1 (en) | Method and apparatus for extracting claim settlement information, and electronic device | |
WO2021114612A1 (en) | Target re-identification method and apparatus, computer device, and storage medium | |
CN110675940A (en) | Pathological image labeling method and device, computer equipment and storage medium | |
CN113239910B (en) | Certificate identification method, device, equipment and storage medium | |
CN111340640A (en) | Insurance claim settlement material auditing method, device and equipment | |
CN111144372A (en) | Vehicle detection method, device, computer equipment and storage medium | |
CN114663871A (en) | Image recognition method, training method, device, system and storage medium | |
CN112991349A (en) | Image processing method, device, equipment and storage medium | |
CN110298302B (en) | Human body target detection method and related equipment |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 18916840 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
32PN | Ep: public notification in the ep bulletin as address of the adressee cannot be established |
Free format text: NOTING OF LOSS OF RIGHTS (EPO FORM 1205A DATED 05.02.2021) |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 18916840 Country of ref document: EP Kind code of ref document: A1 |