WO2019169688A1 - 车辆定损方法、装置、电子设备及存储介质 - Google Patents

车辆定损方法、装置、电子设备及存储介质 Download PDF

Info

Publication number
WO2019169688A1
WO2019169688A1 PCT/CN2018/082577 CN2018082577W WO2019169688A1 WO 2019169688 A1 WO2019169688 A1 WO 2019169688A1 CN 2018082577 W CN2018082577 W CN 2018082577W WO 2019169688 A1 WO2019169688 A1 WO 2019169688A1
Authority
WO
WIPO (PCT)
Prior art keywords
vehicle
picture
damage
loss
confidence
Prior art date
Application number
PCT/CN2018/082577
Other languages
English (en)
French (fr)
Inventor
王健宗
王钰婷
黄章成
肖京
Original Assignee
平安科技（深圳）有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技（深圳）有限公司 filed Critical 平安科技（深圳）有限公司
Publication of WO2019169688A1 publication Critical patent/WO2019169688A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
    • G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
    • G06V20/584—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of vehicle lights or traffic lights
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/20—Administration of product repair or maintenance
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/24323—Tree-organised classifiers
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/08—Insurance

Definitions

  • the present application relates to the field of artificial intelligence, and in particular, to a method and device for determining a vehicle loss, an electronic device, and a storage medium.
  • Artificial intelligence technology has been widely used in various scenarios. For example, in auto insurance, artificial intelligence image recognition technology can be used to judge the degree of deformation and damage of vehicles, reduce the cost of manual surveys, and avoid the risk of human factors.
  • Machine learning in the prior art has been able to help people deal with and solve most of the complex problems. However, in the prior art, the method of machine learning is used, and the recognition accuracy is not high, and there are still major challenges.
  • a method for determining a vehicle loss comprising:
  • the part image with the confidence lower than the threshold is added to the training sample of the vehicle part damage recognition model, and the vehicle part damage recognition model is retrained.
  • the method before the vehicle picture is input into the trained vehicle part segmentation model, before the picture of each part of the vehicle is segmented, the method further includes:
  • the acquired vehicle picture is detected to determine whether the acquired vehicle picture is qualified, and the detected content includes one or more of the following combinations: picture clarity, shooting angle, identifiability of the shooting part, and whether the picture is suspected of tampering;
  • the user is prompted to re-upload the vehicle picture.
  • the method before the image of each part is input to the trained vehicle part damage recognition model, the method further includes:
  • Identify the license plate part and the VIN code part from the vehicle picture identify the license plate number from the license plate part, identify the VIN code from the VIN code part, and use the license plate number or VIN code to identify whether the vehicle is an insured vehicle, when it is an insured vehicle, Determine the degree of damage to the picture of each part of the vehicle.
  • the method further includes prompting the user to wait for the loss result when transmitting the portion of the confidence level below the threshold to the fixed loss person.
  • the method further includes: when the machine loss result of the part is different from the manual damage result, the manual loss result is used as the final loss result of the location where the confidence is lower than the threshold.
  • the method further comprises: determining a machine loss result of the portion where the confidence is higher than the threshold as a final loss result of the portion where the confidence is higher than the threshold.
  • the method further includes:
  • the vehicle picture is input into the trained vehicle identification model, and the brand and model of the vehicle are output;
  • the maintenance cost of the vehicle is calculated based on the maintenance data of each part of the vehicle and transmitted to the user equipment of the user.
  • the method further includes: acquiring the insurance data of the vehicle, determining the claim data according to the insurance data and the maintenance cost, and transmitting the claim data to the device of the user of the vehicle for the user to view.
  • a vehicle loss device comprising:
  • a segmentation module configured to input a vehicle image into the trained vehicle part segmentation model, and segment the picture of each part of the vehicle;
  • the identification module is configured to input the image of each part into the trained vehicle part damage recognition model, and identify the machine loss result of each part and the confidence of outputting the machine damage result of each part;
  • a sending module configured to send a part image with a confidence lower than or equal to a threshold to a user equipment of the loss-receiving person, so that the loss-receiving person performs a loss on the part of the image, and determines a final determination of a location where the confidence is lower than the threshold. Damage result
  • the training module is configured to add a part image with a confidence lower than the threshold to the training sample of the vehicle part damage recognition model, and retrain the vehicle part damage recognition model.
  • An electronic device comprising a memory for storing at least one computer readable instruction, and a processor for executing the at least one computer readable instruction to implement the following steps:
  • the part image with the confidence lower than the threshold is added to the training sample of the vehicle part damage recognition model, and the vehicle part damage recognition model is retrained.
  • the processor is further configured to execute the at least one computer readable instruction to input the vehicle picture into the trained vehicle part segmentation model before segmenting the picture of each part of the vehicle.
  • the acquired vehicle picture is detected to determine whether the acquired vehicle picture is qualified, and the detected content includes one or more of the following combinations: picture clarity, shooting angle, identifiability of the shooting part, and whether the picture is suspected of tampering;
  • the user is prompted to re-upload the vehicle picture.
  • the processor is further configured to execute the at least one computer readable instruction to input the respective part of the picture to the trained vehicle part damage recognition model to implement the following steps:
  • Identify the license plate part and the VIN code part from the vehicle picture identify the license plate number from the license plate part, identify the VIN code from the VIN code part, and use the license plate number or VIN code to identify whether the vehicle is an insured vehicle, when it is an insured vehicle, Determine the degree of damage to the picture of each part of the vehicle.
  • the processor is further configured to execute the at least one computer readable instruction to implement the following steps: when the machine loss result of the part is different from the manual loss result, the manual loss result is The final loss result as the location where the confidence is below the threshold.
  • the processor is further configured to execute the at least one computer readable instruction to implement the step of: determining a machine loss result of a portion having a confidence level higher than a threshold as a portion having a confidence level higher than a threshold The final result of the damage.
  • the processor is further configured to execute the at least one computer readable instruction to implement the following steps:
  • the vehicle picture is input into the trained vehicle identification model, and the brand and model of the vehicle are output;
  • the maintenance cost of the vehicle is calculated based on the maintenance data of each part of the vehicle and transmitted to the user equipment of the user.
  • the processor is further configured to execute the at least one computer readable instruction to: obtain the insurance data of the vehicle, determine the claim data according to the insurance data and the maintenance cost, and send the claim data
  • the device of the user of the vehicle is provided for viewing by the user.
  • a non-transitory readable storage medium storing at least one computer readable instruction, the at least one computer readable instruction being executed by a processor to implement the following steps:
  • the part image with the confidence lower than the threshold is added to the training sample of the vehicle part damage recognition model, and the vehicle part damage recognition model is retrained.
  • the at least one computer readable instruction is executed by the processor before the segmentation of the respective parts of the vehicle is performed, and the following steps are further implemented:
  • the acquired vehicle picture is detected to determine whether the acquired vehicle picture is qualified, and the detected content includes one or more of the following combinations: picture clarity, shooting angle, identifiability of the shooting part, and whether the picture is suspected of tampering;
  • the user is prompted to re-upload the vehicle picture.
  • Identify the license plate part and the VIN code part from the vehicle picture identify the license plate number from the license plate part, identify the VIN code from the VIN code part, and use the license plate number or VIN code to identify whether the vehicle is an insured vehicle, when it is an insured vehicle, Determine the degree of damage to the picture of each part of the vehicle.
  • the following step is further implemented: when the machine loss result of the part is different from the manual loss result, the manual loss result is used as the confidence The final loss result of the part below the threshold.
  • the following step is further implemented: the machine loss result of the location where the confidence is higher than the threshold is used as the final determination of the location where the confidence is higher than the threshold Damage result.
  • the present application obtains a picture of the vehicle; inputs the picture of the vehicle into the trained vehicle part segmentation model, and divides the picture of each part of the vehicle; and inputs the picture of each part into the trained vehicle part damage recognition model Identifying the machine's loss-reduction results for each part and the confidence level of the machine's loss-reduction results for each part of the output; sending a part of the picture with a confidence lower than or equal to the threshold to the user equipment of the fixed-loss person
  • the part image is subjected to the fixed loss, and the final loss result of the portion where the confidence is lower than the threshold is determined; the part image with the confidence lower than the threshold is added to the training sample of the vehicle part damage recognition model, and the vehicle part damage recognition model is retrained.
  • This application can enhance the vehicle part damage recognition model by using the idea of HITL manual intervention, and form the adaptive effect of the model. Through continuous enhancement and update, the recognition accuracy of the vehicle part damage recognition model model is improved.
  • FIG. 1 is a flow chart of a first preferred embodiment of a method for determining a vehicle's loss in the present application.
  • FIG. 2 is a flow chart of a second preferred embodiment of the method for determining the damage of the vehicle of the present application.
  • FIG. 3 is a block diagram of a program of a first preferred embodiment of the vehicle damper device of the present application.
  • FIG. 4 is a block diagram of a program of a second preferred embodiment of the vehicle damper device of the present application.
  • FIG. 5 is a schematic structural diagram of a preferred embodiment of an electronic device in at least one example of the present application.
  • FIG. 1 it is a flow chart of a first preferred embodiment of the method for determining the damage of the vehicle of the present application.
  • the order of the steps in the flowchart may be changed according to different requirements, and some steps may be omitted.
  • the electronic device acquires a picture of the vehicle.
  • the vehicle in the scene of the accident is photographed and the photograph of the photographed vehicle is transmitted to the cloud, the picture of the vehicle including a panoramic picture of the vehicle, a picture of the part of the vehicle, and the like.
  • the acquired vehicle picture is detected, and it is determined whether the acquired vehicle picture is qualified.
  • the acquired vehicle picture is unqualified, the user is prompted to re-upload the vehicle picture.
  • the detected content of the acquired picture includes picture clarity, shooting angle, recognizable degree of the shooting part, whether the picture has tampering suspicion, etc., for example, whether the picture sharpness is within the configured sharpness range, and whether the shooting angle is configured Within the range of angles, whether the identifiable extent of the shot is within the identifiable extent of the configuration, and so on. If the acquired vehicle picture is acceptable, execute S11. In this way, the influence of the unqualified picture on the subsequent damage loss result can be avoided, so as to improve the accuracy of the vehicle damage.
  • the electronic device inputs the vehicle picture into the trained vehicle part segmentation model, and divides the picture of each part of the vehicle.
  • the vehicle part segmentation model is used to segment images of various parts of the vehicle.
  • the training samples of the vehicle part segmentation model include pictures of various parts of the vehicle, such as door handles, doors, tires, and the like.
  • the vehicle part segmentation model continuously learns the features of various parts of the vehicle. After the vehicle part segmentation model is trained, it can be identified from the vehicle picture, and the pictures of various parts of the vehicle are segmented, so as to facilitate the subsequent determination of the damage degree of each part of the picture.
  • the license plate part and the VIN code part are also identified from the vehicle picture
  • the license plate number is identified from the license plate part
  • the VIN code is identified from the VIN code part
  • the vehicle number is used to identify whether the vehicle is an insured vehicle, When the vehicle is insured, the degree of damage of the picture of each part of the vehicle is judged.
  • the training process of the vehicle part segmentation model includes:
  • the electronic device inputs the image of each part into the trained vehicle part damage recognition model, and identifies the machine loss result of each part and the confidence of outputting the machine loss result of each part.
  • the vehicle part damage recognition model is used to determine the degree of damage of each part and output a confidence level of the machine damage result of each part.
  • Training samples for training vehicle part damage identification models include pictures of various degrees of damage at various locations.
  • the vehicle part damage recognition model continuously learns the characteristics of various damage levels of each part. After the vehicle part damage recognition model is trained, it is possible to determine the damage of each part of the picture, determine the degree of damage of each part, and output the confidence of the damage loss results of each part. Subsequent machine loss results with confidence below the threshold can be sent to multiple professionals for loss.
  • the method further comprises: determining a machine loss result of the portion where the confidence is higher than the threshold as a final loss result of the portion where the confidence is higher than the threshold.
  • the electronic device sends a part image with a confidence lower than or equal to the threshold to the user equipment of the loss-receiving person, so that the fixed-loss person performs a loss on the part of the picture, and determines a final part of the location where the confidence is lower than the threshold. The result of the loss.
  • the number of persons to be degraded is one or more.
  • the damage result exceeding the preset number of people is taken as the result of manual damage.
  • a part image with a confidence lower than the threshold is sent to five fixed-loss personnel, four fixed-loss personnel determine the first-level damage degree, and one fixed-loss person determines the second-level damage degree, and the manual damage result is The degree of primary damage.
  • the image of the part with the confidence lower than the threshold is sent to the plurality of fixed-loss personnel to make the damage to the vehicle, and the same loss-reduction result of most of the determined persons is used as the manual damage result, which can effectively avoid the manual. Interference factor.
  • the method further comprises prompting the user to wait for the loss result when transmitting the portion of the confidence level below the threshold to the fixed loss person.
  • the method further comprises: when the machine loss result of the part is different from the manual damage result, the manual loss result is used as the final loss result of the part with a confidence lower than the threshold. In this way, the damage loss result of the vehicle can be manually intervened to improve the accuracy of the loss.
  • the electronic device adds a part image with a confidence lower than a threshold to the training sample of the vehicle part damage recognition model to retrain the vehicle part damage recognition model.
  • the part of the picture with a confidence below the threshold is updated into the training sample of the vehicle part damage recognition model.
  • the model parameters of the secondary damage in the vehicle part damage recognition model need to be strengthened, and it is necessary to learn more secondary parts of the part.
  • the feature of the damage is added to the secondary damage category as a training sample of the secondary damage category. This can increase the sample of the category of the machine learning algorithm to determine the error, and retrain the vehicle part damage recognition model, so that the vehicle part damage recognition model learns the characteristics of the sample of the wrong category, so that the model parameters of the vehicle part damage recognition model are followed. It is possible to accurately determine the degree of vehicle damage of the type of the judgment error.
  • the training process of the vehicle part damage recognition model includes:
  • the vehicle part segmentation model, the vehicle part damage recognition model, and the vehicle type recognition model may be a deep convolutional neural network model without a fully connected layer, the deep convolutional nerve
  • the network model includes an input layer, a convolution layer, a pooling layer, an upsampling layer, and a cutting layer.
  • the deep convolutional neural network model consists of one input layer and 16 convolution layers. , 5 pooling layers, 1 upsampling layer, 1 cutting layer.
  • the recognition model in this embodiment is a deep convolutional neural network model without a fully connected layer.
  • the deep convolutional neural network model only needs to output a classification score for each pixel on a Conv8 layer. .
  • each point on the feature map has scores of different classifications in class num+1 categories, so the output channel is also class num+1, and the recognition efficiency is greatly improved.
  • the present application can add a part image with a confidence lower than the threshold to the training sample of the vehicle part damage recognition model to retrain the vehicle part damage recognition model, so that the vehicle part damage recognition model can be intensively learned, and the model is adaptive.
  • the effect is to improve the recognition accuracy of the vehicle part damage recognition model model by continuously strengthening and updating.
  • FIG. 2 is a flow chart of a second preferred embodiment of the method for determining the damage of the vehicle of the present application.
  • the order of the steps in the flowchart may be changed according to different requirements, and some steps may be omitted.
  • S20 to S24 correspond to S10 to S14 in the first preferred embodiment, respectively, and will not be described in detail herein.
  • the electronic device inputs the vehicle picture into the trained vehicle identification model, and outputs the brand and the vehicle type of the vehicle.
  • a panoramic picture of the vehicle picture is used as an input to the vehicle type recognition model.
  • the training sample of the vehicle identification model is a panoramic picture of various models of various brand vehicles.
  • the vehicle identification model is trained, the brand and model of the vehicle in the input panoramic picture can be automatically recognized.
  • the training algorithm is prior art, including, but not limited to, a convolutional neural network model.
  • the training process in the vehicle type recognition model includes:
  • the electronic device determines maintenance data of each part of the vehicle according to a brand and a model of the vehicle, and a final loss result of each part of the vehicle.
  • the service data includes price data for repair parts, maintenance work cost data.
  • the different degrees of damage of each part correspond to different repair parts and different maintenance man-hours.
  • the maintenance data of each part of the vehicle of the brand and model is read from the database.
  • an inquiry command is sent to the supplier to inquire about maintenance data of various parts of the vehicle of the brand and model.
  • the electronic device calculates a maintenance cost of the vehicle according to the maintenance data of each part of the vehicle, and sends the repair cost to the user equipment.
  • the price data of the repair parts of the various parts of the vehicle and the maintenance work cost data are accumulated as the maintenance cost of the vehicle.
  • the method further includes: acquiring the insurance data of the vehicle, determining the claim data according to the insurance data and the maintenance cost, and transmitting the claim data to the device of the user of the vehicle for the user to view.
  • the present application can accurately damage the damaged vehicle, and calculate the maintenance cost and claim data of the vehicle according to the fixed damage condition and the insurance data of the vehicle, thereby improving the efficiency of claim settlement and the transparency of the claim.
  • FIG. 3 is a block diagram showing the program of the first preferred embodiment of the vehicle loss device of the present application.
  • the vehicle damage device 3 includes, but is not limited to, one or more of the following modules: an acquisition module 30, a segmentation module 31, an identification module 32, a training module 33, a determination module 34, a transmission module 35, and a prompt module 36.
  • a unit referred to in this application refers to a series of computer program segments that can be executed by a processor of the vehicle's loss-limiting device 3 and that are capable of performing a fixed function, which is stored in a memory. The function of each unit will be detailed in the subsequent embodiments.
  • the acquisition module 30 acquires a vehicle picture.
  • the vehicle in the scene of the accident is photographed and the photograph of the photographed vehicle is transmitted to the cloud, the picture of the vehicle including a panoramic picture of the vehicle, a picture of the part of the vehicle, and the like.
  • the acquiring module 30 is further configured to: detect the acquired vehicle image, and determine whether the acquired vehicle image is qualified.
  • the acquired vehicle image is unqualified, the user is prompted to re-upload the vehicle image.
  • the detected content of the acquired picture includes picture clarity, shooting angle, recognizable degree of the shooting part, whether the picture has tampering suspicion, etc., for example, whether the picture sharpness is within the configured sharpness range, and whether the shooting angle is configured Within the range of angles, whether the identifiable extent of the shot is within the identifiable extent of the configuration, and so on. If the acquired vehicle picture is acceptable, the segmentation module 31 is executed. In this way, the influence of the unqualified picture on the subsequent damage loss result can be avoided, so as to improve the accuracy of the vehicle damage.
  • the segmentation module 31 inputs the vehicle picture into the trained vehicle part segmentation model, and segments the picture of each part of the vehicle.
  • the vehicle part segmentation model is used to segment images of various parts of the vehicle.
  • the training samples of the vehicle part segmentation model include pictures of various parts of the vehicle, such as door handles, doors, tires, and the like.
  • the vehicle part segmentation model continuously learns the features of various parts of the vehicle. After the vehicle part segmentation model is trained, it can be identified from the vehicle picture, and the pictures of various parts of the vehicle are segmented, so as to facilitate the subsequent determination of the damage degree of each part of the picture.
  • the segmentation module 31 is further configured to: identify a license plate part, a VIN code part from a vehicle picture, identify a license plate number from the license plate part, identify a VIN code from the VIN code part, and identify with a license plate number or a VIN code. Whether the vehicle is an insured vehicle or not, when it is an insured vehicle, it determines the degree of damage of the picture of each part of the vehicle.
  • the training process of the training module 33 in the vehicle part segmentation model includes:
  • the electronic component device of the identification module 32 inputs the image of each part into the trained vehicle part damage recognition model, and identifies the machine loss result of each part and the confidence of outputting the machine loss result of each part.
  • the vehicle part damage recognition model is used to determine the degree of damage of each part and output a confidence level of the machine damage result of each part.
  • Training samples for training vehicle part damage identification models include pictures of various degrees of damage at various locations.
  • the vehicle part damage recognition model continuously learns the characteristics of various damage levels of each part. After the vehicle part damage recognition model is trained, it is possible to determine the damage of each part of the picture, determine the degree of damage of each part, and output the confidence of the damage loss results of each part. Subsequent machine loss results with confidence below the threshold can be sent to multiple professionals for loss.
  • the determining module 34 is configured to: use a machine loss result of a portion where the confidence is higher than the threshold as a final loss result of a portion where the confidence is higher than the threshold.
  • the sending module 35 sends a part image with a confidence lower than or equal to the threshold to the user equipment of the loss-receiving person, so that the fixed-loss person performs the loss on the part picture, and determines the final determination of the part with the confidence lower than the threshold. Damage result.
  • the number of persons to be degraded is one or more.
  • the damage result exceeding the preset number of people is taken as the result of manual damage.
  • a part image with a confidence lower than the threshold is sent to five fixed-loss personnel, four fixed-loss personnel determine the first-level damage degree, and one fixed-loss person determines the second-level damage degree, and the manual damage result is The degree of primary damage.
  • the image of the part with the confidence lower than the threshold is sent to the plurality of fixed-loss personnel to make the damage to the vehicle, and the same loss-reduction result of most of the determined persons is used as the manual damage result, which can effectively avoid the manual. Interference factor.
  • the prompting module 36 is configured to prompt the user to wait for the damage loss result when the location of the location with the confidence lower than the threshold is sent to the determined loss person.
  • the determining module 34 is configured to: when the machine loss result of the part is different from the manual damage result, use the manual loss result as the final loss result of the location where the confidence is lower than the threshold. In this way, the damage loss result of the vehicle can be manually intervened to improve the accuracy of the loss.
  • the training module 33 adds a part image with a confidence lower than the threshold to the training sample of the vehicle part damage recognition model to retrain the vehicle part damage recognition model.
  • the training module 33 updates the part image with a confidence lower than the threshold to the training sample of the vehicle part damage recognition model. For example, if the machine damage result of the part image is secondary damage and the manual damage result is first-order damage, the model parameters of the secondary damage in the vehicle part damage recognition model need to be strengthened, and it is necessary to learn more secondary parts of the part.
  • the feature of the damage is added to the secondary damage category as a training sample of the secondary damage category. This can increase the sample of the category of the machine learning algorithm to determine the error, and retrain the vehicle part damage recognition model, so that the vehicle part damage recognition model learns the characteristics of the sample of the wrong category, so that the model parameters of the vehicle part damage recognition model are followed. It is possible to accurately determine the degree of vehicle damage of the type of the judgment error.
  • the training process of the training module 33 in the vehicle part damage recognition model includes:
  • the vehicle part segmentation model, the vehicle part damage recognition model, and the vehicle type recognition model may be a deep convolutional neural network model without a fully connected layer, the deep convolutional nerve
  • the network model includes an input layer, a convolution layer, a pooling layer, an upsampling layer, and a cutting layer.
  • the deep convolutional neural network model consists of one input layer and 16 convolution layers. , 5 pooling layers, 1 upsampling layer, 1 cutting layer.
  • the recognition model in this embodiment is a deep convolutional neural network model without a fully connected layer.
  • the deep convolutional neural network model only needs to output a classification score for each pixel on a Conv8 layer. .
  • each point on the feature map has scores of different classifications in class num+1 categories, so the output channel is also class num+1, and the recognition efficiency is greatly improved.
  • the present application can add a part image with a confidence lower than the threshold to the training sample of the vehicle part damage recognition model to retrain the vehicle part damage recognition model, so that the vehicle part damage recognition model can be intensively learned, and the model is adaptive.
  • the effect is to improve the recognition accuracy of the vehicle part damage recognition model model by continuously strengthening and updating.
  • the vehicle damage device 3 includes one or more modules in the first preferred embodiment: an acquisition module 30, a segmentation module 31, an identification module 32, a training module 33, a determination module 34, a transmission module 35, and a prompt module 36.
  • the vehicle damage device 3 may further include one or more of the following modules: an output module 37 and a calculation module 38.
  • a unit referred to in this application refers to a series of computer program segments that can be executed by a processor of the vehicle's loss-limiting device 3 and that are capable of performing a fixed function, which is stored in a memory. The function of each unit will be detailed in the subsequent embodiments.
  • the output module 37 inputs the vehicle picture into the trained vehicle identification model to output the brand and vehicle type of the vehicle.
  • a panoramic picture of the vehicle picture is used as an input to the vehicle type recognition model.
  • the training sample of the vehicle identification model is a panoramic picture of various models of various brand vehicles.
  • the vehicle identification model is trained, the brand and model of the vehicle in the input panoramic picture can be automatically recognized.
  • the training algorithm is prior art, including, but not limited to, a convolutional neural network model.
  • the training process of the training module 33 in the vehicle type recognition model includes:
  • the determining module 34 determines maintenance data for various parts of the vehicle based on the brand and model of the vehicle and the final loss results of the various parts of the vehicle.
  • the service data includes price data for repair parts, maintenance work cost data.
  • the different degrees of damage of each part correspond to different repair parts and different maintenance man-hours.
  • the maintenance data of each part of the vehicle of the brand and model is read from the database.
  • an inquiry command is sent to the supplier to inquire about maintenance data of various parts of the vehicle of the brand and model.
  • the calculation module 38 calculates the maintenance cost of the vehicle according to the maintenance data of each part of the vehicle, and transmits it to the user equipment of the user.
  • the price data of the repair parts of the various parts of the vehicle and the maintenance work cost data are accumulated as the maintenance cost of the vehicle.
  • the determining module 34 is further configured to: obtain the insurance data of the vehicle, determine the claim data according to the insurance data and the maintenance cost, and send the claim data to the device of the user of the vehicle for the user to view.
  • the present application can accurately damage the damaged vehicle, and calculate the maintenance cost and claim data of the vehicle according to the fixed damage condition and the insurance data of the vehicle, thereby improving the efficiency of claim settlement and the transparency of the claim.
  • the above-described integrated unit implemented in the form of a software program module can be stored in a computer readable storage medium.
  • the software program module described above is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor to perform the method of each embodiment of the present application. Part of the steps.
  • the electronic device 5 comprises at least one transmitting device 51, at least one memory 52, at least one processor 53, at least one receiving device 54, and at least one communication bus.
  • the communication bus is used to implement connection communication between these components.
  • the electronic device 5 is a device capable of automatically performing numerical calculation and/or information processing according to an instruction set or stored in advance, and the hardware includes, but not limited to, a microprocessor, an application specific integrated circuit (ASIC). ), Field-Programmable Gate Array (FPGA), Digital Signal Processor (DSP), embedded devices, etc.
  • the electronic device 5 may also comprise a network device and/or a user device.
  • the network device includes, but is not limited to, a single network server, a server group composed of multiple network servers, or a cloud computing-based cloud composed of a large number of hosts or network servers, where the cloud computing is distributed computing.
  • a super virtual computer consisting of a group of loosely coupled computers.
  • the electronic device 5 can be, but is not limited to, any electronic product that can interact with a user through a keyboard, a touch pad, or a voice control device, such as a tablet, a smart phone, or a personal digital assistant (Personal Digital Assistant). , PDA), smart wearable devices, camera equipment, monitoring equipment and other terminals.
  • a keyboard e.g., a keyboard
  • a touch pad e.g., a touch pad
  • a voice control device such as a tablet, a smart phone, or a personal digital assistant (Personal Digital Assistant). , PDA), smart wearable devices, camera equipment, monitoring equipment and other terminals.
  • PDA Personal Digital Assistant
  • the network in which the electronic device 5 is located includes, but is not limited to, the Internet, a wide area network, a metropolitan area network, a local area network, a virtual private network (VPN), and the like.
  • the Internet includes, but is not limited to, the Internet, a wide area network, a metropolitan area network, a local area network, a virtual private network (VPN), and the like.
  • VPN virtual private network
  • the receiving device 54 and the sending device 51 may be wired transmission ports, or may be wireless devices, for example, including antenna devices, for performing data communication with other devices.
  • the memory 52 is used to store program code.
  • the memory 52 may be a circuit having a storage function, such as a RAM (Random-Access Memory), a FIFO (First In First Out), or the like, which is not in a physical form in the integrated circuit.
  • the memory 52 may also be a memory having a physical form, such as a memory stick, a TF card (Trans-flash Card), a smart media card, a secure digital card, a flash memory card.
  • Storage devices such as (flash card) and the like.
  • the processor 53 can include one or more microprocessors, digital processors.
  • the processor 53 can call program code stored in the memory 52 to perform related functions.
  • the various modules described in Figures 3 and 4 are program code stored in the memory 52 and executed by the processor 53 to implement a method of vehicle loss.
  • the processor 53 also known as a central processing unit (CPU), is a very large-scale integrated circuit, which is a computing core (Core) and a control unit (Control Unit).
  • the embodiment of the present application further provides a computer readable storage medium having stored thereon computer instructions, when executed by an electronic device including one or more processors, causing the electronic device to perform the method embodiment as described above Vehicle loss method.
  • the memory 52 in the electronic device 5 stores a plurality of instructions to implement a vehicle loss method
  • the processor 53 can execute the plurality of instructions to implement:
  • the processor executes the plurality of instructions: when the vehicle picture is input into the trained vehicle part segmentation model, before the pictures of the respective parts of the vehicle are segmented, the acquired vehicle picture is detected, and the obtained is determined. Whether the vehicle image is qualified or not, the detected content includes one or more of the following combinations: picture clarity, shooting angle, identifiability of the shooting location, and whether the picture is suspected of tampering;
  • the user is prompted to re-upload the vehicle picture.
  • the processor executes the plurality of instructions: before inputting the image of each part to the trained vehicle part damage recognition model, identifying the license plate part, the VIN code part, and the license plate part from the vehicle picture
  • the license plate number is recognized
  • the VIN code is recognized from the VIN code part
  • the vehicle number or VIN code is used to identify whether the vehicle is an insured vehicle.
  • the damage degree of each part of the vehicle is determined.
  • the processor executes the plurality of instructions, the following instructions are further executed: when transmitting the portion of the confidence level below the threshold to the fixed-loss person, prompting the user to wait for the loss-making result.
  • the processor executes the plurality of instructions, the following instruction is further executed: when the machine loss result of the part is different from the manual damage result, the manual loss result is used as the final loss result of the part with a confidence lower than the threshold. .
  • the processor executes the plurality of instructions, the following instructions are further executed: the machine loss result of the portion where the confidence is higher than the threshold is the final loss result of the portion where the confidence is higher than the threshold.
  • the processor also executes the following instructions when executing the plurality of instructions:
  • the vehicle picture is input into the trained vehicle identification model, and the brand and model of the vehicle are output;
  • the maintenance cost of the vehicle is calculated based on the maintenance data of each part of the vehicle and transmitted to the user equipment of the user.
  • the processor further executes the following instructions: acquiring the insurance data of the vehicle, determining the claim data according to the insurance data and the maintenance cost, and transmitting the claim data to the user's device of the vehicle for the user to view.
  • the plurality of instructions corresponding to the vehicle loss method are stored in the memory 52 in any of the embodiments and executed by the processor 53, and are not described in detail herein.
  • the above-described characteristic means of the present application can be implemented by an integrated circuit and control the function of implementing the vehicle loss-reduction method described in any of the above embodiments. That is, the integrated circuit of the present application is installed in the electronic device, so that the electronic device performs the following functions: acquiring a picture of the vehicle; inputting the picture of the vehicle into the trained vehicle part segmentation model, and segmenting the picture of each part of the vehicle; Input the image of each part into the trained vehicle part damage recognition model, identify the machine loss result of each part and the confidence of the machine loss result of each part; send the part image with the confidence lower than or equal to the threshold to
  • the user equipment of the loss-receiving personnel is configured to cause the loss-receiving personnel to determine the loss of the image of the part, and determine the final loss result of the part where the confidence is lower than the threshold; and add the image of the part whose confidence is lower than the threshold to the damage identification of the vehicle part
  • the vehicle part damage recognition model is retrained.
  • the functions that can be implemented by the vehicle loss-receiving method in any of the embodiments can be installed in the electronic device through the integrated circuit of the present application, so that the electronic device can perform the method of determining the damage of the vehicle in any of the embodiments.
  • the functions implemented are not detailed here.
  • the disclosed apparatus may be implemented in other ways.
  • the device embodiments described above are merely illustrative.
  • the division of the unit is only a logical function division.
  • there may be another division manner for example, multiple units or components may be combined or may be Integrate into another system, or some features can be ignored or not executed.
  • the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, and may be electrical or otherwise.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.
  • each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
  • the above integrated unit can be implemented in the form of hardware or in the form of a software functional unit.
  • the integrated unit if implemented in the form of a software functional unit and sold or used as a standalone product, may be stored in a computer readable storage medium.
  • a computer readable storage medium A number of instructions are included to cause a computer device (which may be a personal computer, server or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present application.
  • the foregoing storage medium includes: a U disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic disk, or an optical disk, and the like. .

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Strategic Management (AREA)
  • Accounting & Taxation (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Finance (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Economics (AREA)
  • Human Resources & Organizations (AREA)
  • Molecular Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Operations Research (AREA)
  • Tourism & Hospitality (AREA)
  • Quality & Reliability (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Development Economics (AREA)
  • Technology Law (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)
  • Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)

Abstract

本申请提供一种车辆定损方法，所述方法包括：获取车辆图片；将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片；将各个部位图片输入至训练好的车辆部位损伤识别模型中，识别各个部位的机器定损结果及输出各个部位的机器定损结果的置信度；将置信度低于或等于阈值的部位图片发送至定损人员的用户设备上以使定损人员对该部位图片进行定损，并确定置信度低于阈值的部位的最终定损结果；将置信度低于阈值的部位图片添加至车辆部位损伤识别模型的训练样本中，重新训练车辆部位损伤识别模型。本申请还提供一种电子设备及存储介质。本申请能使车辆部位损伤识别模型得到强化学习，通过不断强化和更新，提高模型识别准确率。

Description

车辆定损方法、装置、电子设备及存储介质
本申请要求于2018年03月09日提交中国专利局，申请号为201810196561.1发明名称为“车辆定损方法、装置、电子设备及存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及人工智能领域，尤其涉及一种车辆定损方法、装置、电子设备及存储介质。
背景技术
人工智能技术目前已经广泛应用在各种场景当中，例如在车险中，可以通过人工智能图像识别技术来判断车辆的变形和损伤程度，减少人工勘察成本，规避人为因素风险。现有技术中的机器学习已经能够帮助人们处理和解决大部分的复杂问题。但现有技术中利用机器学习的方法，识别精度不高，仍然存在着较大的挑战。
发明内容
鉴于以上内容，有必要提供一种车辆定损方法、装置、电子设备及存储介质，能通过采用HITL人工介入的思想，使车辆部位损伤识别模型得到强化学习，形成模型的自适应效果，通过不断强化和更新，提高车辆部位损伤识别模型模型识别准确率。
一种车辆定损方法，所述方法包括：
获取车辆图片；
将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片；
将各个部位图片输入至训练好的车辆部位损伤识别模型中，识别各个部位的机器定损结果及输出各个部位的机器定损结果的置信度；
将置信度低于或等于阈值的部位图片发送至定损人员的用户设备上以使定损人员对该部位图片进行定损，并确定置信度低于阈值的部位的最终定损结果；
将置信度低于阈值的部位图片添加至车辆部位损伤识别模型的训练样本中，重新训练车辆部位损伤识别模型。
根据本申请优选实施例，在将车辆图片输入至训练好的车辆部位分割模型 中，分割出车辆的各个部位图片之前，所述方法还包括：
对获取的车辆图片进行检测，判断获取的车辆图片是否合格，检测的内容包括以下一种或者多种的组合：图片清晰度、拍摄角度、拍摄部位的可识别程度、图片是否存在篡改嫌疑；
当所述获取的车辆图片不合格时，提示用户重新上传车辆图片。
根据本申请优选实施例，在将各个部位图片输入至训练好的车辆部位损伤识别模型之前，所述方法还包括：
从车辆图片中识别出车牌部位、VIN码部位，从车牌部位中识别出车牌号，从VIN码部位识别出VIN码，利用车牌号或者VIN码识别车辆是否为投保车辆，当为投保车辆时，判断车辆的各个部位图片的损伤程度。
根据本申请优选实施例，所述方法还包括：当将置信度低于阈值的部位的发送至定损人员时，提示用户等待定损结果。
根据本申请优选实施例，所述方法还包括：当该部位的机器定损结果与人工定损结果不同时，将人工定损结果作为置信度低于阈值的部位的最终定损结果。
根据本申请优选实施例，所述方法还包括：将置信度高于阈值的部位的机器定损结果作为置信度高于阈值的部位的最终定损结果。
根据本申请优选实施例，所述方法还包括：
当车辆为投保车辆时，将所述车辆图片输入至训练好的车型识别模型中，输出车辆的品牌和车型；
根据车辆的品牌和车型,及车辆的各个部位的最终定损结果，确定车辆的各个部位的维修数据；
根据车辆的各个部位的维修数据，计算车辆的维修费用，并发送给用户的用户设备。
根据本申请优选实施例，所述方法还包括：获取车辆的投保数据，根据投保数据及维修费用，确定理赔数据，将理赔数据发送给该车辆的用户的设备以供用户查看。
一种车辆定损装置，所述装置包括：
获取模块，用于获取车辆图片；
分割模块，用于将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片；
识别模块，用于将各个部位图片输入至训练好的车辆部位损伤识别模型中，识别各个部位的机器定损结果及输出各个部位的机器定损结果的置信度；
发送模块，用于将置信度低于或等于阈值的部位图片发送至定损人员的用户设备上以使定损人员对该部位图片进行定损，并确定置信度低于阈值的部位的最终定损结果；
训练模块，用于将置信度低于阈值的部位图片添加至车辆部位损伤识别模 型的训练样本中，重新训练车辆部位损伤识别模型。
一种电子设备，所述电子设备包括存储器及处理器，所述存储器用于存储至少一个计算机可读指令，所述处理器用于执行所述至少一个计算机可读指令以实现以下步骤：
获取车辆图片；
将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片；
将各个部位图片输入至训练好的车辆部位损伤识别模型中，识别各个部位的机器定损结果及输出各个部位的机器定损结果的置信度；
将置信度低于或等于阈值的部位图片发送至定损人员的用户设备上以使定损人员对该部位图片进行定损，并确定置信度低于阈值的部位的最终定损结果；
将置信度低于阈值的部位图片添加至车辆部位损伤识别模型的训练样本中，重新训练车辆部位损伤识别模型。
根据本申请优选实施例，在将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片之前，所述处理器还用于执行所述至少一个计算机可读指令，以实现以下步骤：
对获取的车辆图片进行检测，判断获取的车辆图片是否合格，检测的内容包括以下一种或者多种的组合：图片清晰度、拍摄角度、拍摄部位的可识别程度、图片是否存在篡改嫌疑；
当所述获取的车辆图片不合格时，提示用户重新上传车辆图片。
根据本申请优选实施例，在将各个部位图片输入至训练好的车辆部位损伤识别模型之前，所述处理器还用于执行所述至少一个计算机可读指令，以实现以下步骤：
从车辆图片中识别出车牌部位、VIN码部位，从车牌部位中识别出车牌号，从VIN码部位识别出VIN码，利用车牌号或者VIN码识别车辆是否为投保车辆，当为投保车辆时，判断车辆的各个部位图片的损伤程度。
根据本申请优选实施例，所述处理器还用于执行所述至少一个计算机可读指令，以实现以下步骤：当该部位的机器定损结果与人工定损结果不同时，将人工定损结果作为置信度低于阈值的部位的最终定损结果。
根据本申请优选实施例，所述处理器还用于执行所述至少一个计算机可读指令，以实现以下步骤：将置信度高于阈值的部位的机器定损结果作为置信度高于阈值的部位的最终定损结果。
根据本申请优选实施例，所述处理器还用于执行所述至少一个计算机可读指令，以实现以下步骤：
当车辆为投保车辆时，将所述车辆图片输入至训练好的车型识别模型中，输出车辆的品牌和车型；
根据车辆的品牌和车型，及车辆的各个部位的最终定损结果，确定车辆的各个部位的维修数据；
根据车辆的各个部位的维修数据，计算车辆的维修费用，并发送给用户的用户设备。
根据本申请优选实施例，所述处理器还用于执行所述至少一个计算机可读指令，以实现以下步骤：获取车辆的投保数据，根据投保数据及维修费用，确定理赔数据，将理赔数据发送给该车辆的用户的设备以供用户查看。
一种非易失性可读存储介质，所述计算机可读存储介质存储有至少一个计算机可读指令，所述至少一个计算机可读指令被处理器执行时实现以下步骤：
获取车辆图片；
将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片；
将各个部位图片输入至训练好的车辆部位损伤识别模型中，识别各个部位的机器定损结果及输出各个部位的机器定损结果的置信度；
将置信度低于或等于阈值的部位图片发送至定损人员的用户设备上以使定损人员对该部位图片进行定损，并确定置信度低于阈值的部位的最终定损结果；
将置信度低于阈值的部位图片添加至车辆部位损伤识别模型的训练样本中，重新训练车辆部位损伤识别模型。
根据本申请优选实施例，在将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片之前，所述至少一个计算机可读指令被处理器执行时，还实现以下步骤：
对获取的车辆图片进行检测，判断获取的车辆图片是否合格，检测的内容包括以下一种或者多种的组合：图片清晰度、拍摄角度、拍摄部位的可识别程度、图片是否存在篡改嫌疑；
当所述获取的车辆图片不合格时，提示用户重新上传车辆图片。
根据本申请优选实施例，在将各个部位图片输入至训练好的车辆部位损伤识别模型之前，所述至少一个计算机可读指令被处理器执行时，还实现以下步骤：
从车辆图片中识别出车牌部位、VIN码部位，从车牌部位中识别出车牌号，从VIN码部位识别出VIN码，利用车牌号或者VIN码识别车辆是否为投保车辆，当为投保车辆时，判断车辆的各个部位图片的损伤程度。
根据本申请优选实施例，所述至少一个计算机可读指令被处理器执行时，还实现以下步骤：当该部位的机器定损结果与人工定损结果不同时，将人工定损结果作为置信度低于阈值的部位的最终定损结果。
根据本申请优选实施例，所述至少一个计算机可读指令被处理器执行时，还实现以下步骤：将置信度高于阈值的部位的机器定损结果作为置信度高于阈值的部位的最终定损结果。
由以上技术方案可以看出，本申请获取车辆图片；将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片；将各个部位图片输入至训练好的车辆部位损伤识别模型中，识别各个部位的机器定损结果及输出各个部位的机器定损结果的置信度；将置信度低于或等于阈值的部位图片发送至定损人员的用户设备上以使定损人员对该部位图片进行定损，并确定置信度低于阈值的部位的最终定损结果；将置信度低于阈值的部位图片添加至车辆部位损伤识别模型的训练样本中，重新训练车辆部位损伤识别模型。本申请能通过采用 HITL人工介入的思想，使车辆部位损伤识别模型得到强化学习，形成模型的自适应效果，通过不断强化和更新，提高车辆部位损伤识别模型模型识别准确率。
附图说明
为了更清楚地说明本申请实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据提供的附图获得其他的附图。
图1是本申请车辆定损方法的第一较佳实施例的流程图。
图2是本申请车辆定损方法的第二较佳实施例的流程图。
图3是本申请车辆定损装置的第一较佳实施例的程序模块图。
图4是本申请车辆定损装置的第二较佳实施例的程序模块图。
图5是本申请至少一个实例中电子设备的较佳实施例的结构示意图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
为使本申请的上述目的、特征和优点能够更加明显易懂，下面结合附图和具体实施方式对本申请作进一步详细的说明。
为了使本技术领域的人员更好地理解本申请方案，下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本申请一部分的实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都应当属于本申请保护的范围。
本申请的说明书和权利要求书及上述附图中的术语“第一”、“第二”和“第三”等是用于区别不同对象，而非用于描述特定顺序。此外，术语“包括”以及它们任何变形，意图在于覆盖不排他的包含。例如包含了一系列步骤或单元的过程、方法、系统、产品或设备没有限定于已列出的步骤或单元，而是可选地还包括没有列出的步骤或单元，或可选地还包括对于这些过程、方法、产品或设备固有的其它步骤或单元。
如图1所示，是本申请车辆定损方法的第一较佳实施例的流程图。根据不同的需求，该流程图中步骤的顺序可以改变，某些步骤可以省略。
S10、电子设备获取车辆图片。
在可选实施例中，对事故现场中的车辆进行拍照，将拍摄的车辆图片发送至云端，所述车辆图片包括车辆全景图片、车辆部分图片等等。
优选地，对获取的车辆图片进行检测，并判断获取的车辆图片是否合格，当所述获取的车辆图片不合格时，提示用户重新上传车辆图片。对所述获取的图片的检测内容包含图片清晰度、拍摄角度、拍摄部位的可识别程度、图片是否存在篡改嫌疑等，例如图片清晰度是否在配置的清晰度范围内，拍摄角度是否在配置的角度范围内、拍摄部位的可识别程度是否在配置的可识别程度范围内等等。若获取的车辆图片合格时，执行S11。这样可以避免不合格的图片对后续的定损结果造成的影响，以提高车辆定损的准确度。
S11、所述电子设备将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片。
在可选实施中，所述车辆部位分割模型用于分割车辆的各个部位图片。所述车辆部位分割模型的训练样本包括车辆的各个部位的图片，如门把手，车门，轮胎，等等。在训练车辆部位分割模型的过程中，车辆部位分割模型不断学习车辆的各个部位的特征。当车辆部位分割模型训练好后，可以从车辆图片中识别，并分割出车辆的各个部位图片，便于后续判断各个部位图片的损伤程度。
优选地，还从车辆图片中识别出车牌部位、VIN码部位，从车牌部位中识别出车牌号，从VIN码部位识别出VIN码，利用车牌号或者VIN码识别车辆是否为投保车辆，当为投保车辆时，判断车辆的各个部位图片的损伤程度。
可选地，在所述车辆部位分割模型的训练过程包括：
A、配置各个部位(例如，门把手，车门，轮胎、左前门、右前门、左叶子板、右叶子板、前保险杠、后保险杠等)对应的预设数量的样本图片；
B、将各个样本图片进行图片预处理以获得训练所述车辆部位分割模型训练图片，例如可通过对各个样本图片进行图片预处理如缩放、裁剪、翻转及/或扭曲等操作后，使训练图片具有相同的尺寸及相同的视角后，才进行模型训练，以有效提高模型训练的真实性及准确率。
C、将所有训练图片分为第一比例(例如，70％)的训练集、第二比例(例如，30％)的验证集；
D、利用所述训练集训练所述车辆部位分割模型；
E、利用所述验证集验证训练的车辆部位分割模型的准确率，若准确率大于或者等于预设准确率，则训练结束，或者，若准确率小于预设准确率，则增加各个部位对应的样本图片
数量并重新执行上述步骤B、C、D、E，直至训练的车辆部位分割模型的准确率大于或者等于预设准确率。
S12、所述电子设备将各个部位图片输入至训练好的车辆部位损伤识别模型中，识别各个部位的机器定损结果及输出各个部位的机器定损结果的置信度。
在可选实施例中，所述车辆部位损伤识别模型用于判断各个部位的损伤程度，并输出各个部位的机器定损结果的置信度。训练车辆部位损伤识别模型的训练样本包括各个部位的各种损伤程度的图片。在训练车辆部位损伤识别模型的过程中，车辆部位损伤识别模型不断学习各个部位的各种损伤程度的特征。当车辆部位损伤识别模型训练好后，能对各个部位图片进行定损，判定各个部位的损伤程度，并输出各个部位的定损结果的置信度。后续可以将置信度低于阈值的机器定损结果发给多个专业人员进行定损。
优选地，所述方法还包括：将置信度高于阈值的部位的机器定损结果作为置信度高于阈值的部位的最终定损结果。
S13、所述电子设备将置信度低于或等于阈值的部位图片发送至定损人员的用户设备上以使定损人员对该部位图片进行定损，并确定置信度低于阈值的部位的最终定损结果。
在可选实施例中，定损人员为一个或者多个。将超过预设人数的定损结果作为人工定损结果。例如，将置信度低于阈值的部位图片发送至5个定损人员，有4个定损人员判定为一级损伤程度，1个定损人员判定为二级损伤程度，则人工定损结果为一级损伤程度。这样将置信度低于阈值的部位图片发送给多个定损人员以使定损人员对车辆定损，并采用大部分定损人员的相同的定损结果作为人工定损结果，可以有效避免人工干扰因素。
优选地，所述方法还包括：当将置信度低于阈值的部位的发送至定损人员时，提示用户等待定损结果。
优选地，所述方法还包括：当该部位的机器定损结果与人工定损结果不同时，将人工定损结果作为置信度低于阈值的部位的最终定损结果。这样可以人工干预车辆的定损结果，提高定损精度。
S14、所述电子设备将置信度低于阈值的部位图片添加至车辆部位损伤识别模型的训练样本中重新训练车辆部位损伤识别模型。
在可选实施例中，将该置信度低于阈值的部位图片更新至车辆部位损伤识别模型的训练样本中。例如，该部位图片的机器定损结果为二级损伤，人工定损结果为一级损伤，则车辆部位损伤识别模型中二级损伤的模型参数需要强化，需要学习更多的该部位的二级损伤的特征，将置信度低于阈值的部位图片添加至二级损伤类别中，作为二级损伤类别的训练样本。这样可以增加机器学习算法判定错误的类别的样本，并对车辆部位损伤识别模型重新训练，使车辆部位损伤识别模型学习判定错误的类别的样本的特征，从而使车辆部位损伤识别模型的模型参数后续能准确的判定该判断错误的类别的车辆 损伤程度。
可选地，在所述车辆部位损伤识别模型的训练过程包括：
A、配置各个部位的各种损伤程度(例如，对于门把手部位，分别配置一级损伤程度、二级损伤程度、三级损伤程度等等各种损伤程度)对应的预设数量的样本图片；
B、将各个样本图片进行图片预处理以获得训练所述车辆部位损伤识别模型的训练图片，例如可通过对各个样本图片进行图片预处理如缩放、裁剪、翻转及/或扭曲等操作后，使训练图片具有相同的尺寸及相同的视角后，才进行模型训练，以有效提高模型训练的真实性及准确率。
C、将所有训练图片分为第一比例(例如，80％)的训练集、第二比例(例如，20％)的验证集；
D、利用所述训练集训练所述车辆部位损伤识别模型；
E、利用所述验证集验证训练的车辆部位损伤识别模型的准确率，若准确率大于或者等于预设准确率，则训练结束，或者，若准确率小于预设准确率，则增加各个部位对应的样本图片；
数量并重新执行上述步骤B、C、D、E，直至训练的车辆部位损伤识别模型的准确率大于或者等于预设准确率。
在可选实施例中，所述车辆部位分割模型、所述车辆部位损伤识别模型及所述车型识别模型可以是为不带有全连接层的深度卷积神经网络模型，所述深度卷积神经网络模型包括输入层、卷积层、池化层、上采样层及裁切层，在一种具体的实施方式中，所述深度卷积神经网络模型由1个输入层，16个卷积层，5个池化层，1个上采样层，1个裁切层构成。
由于在传统的分类问题中，一般需要用全连接层来输出每一张图片属于每个类的概率，然而在语义分割问题上，用这种方法来预测每个样本属于哪个类必然会导致效率低下。因此，本实施例中的识别模型为不带有全连接层的深度卷积神经网络模型，该深度卷积神经网络模型只需在Conv8上，用一个卷积层来输出每个像素的分类score。在该层上，特征图上的每个点都有class num+1个分类中不同分类的score，因此输出的channel也是class num+1，识别效率大大提高。
通过上述实施，本申请能将置信度低于阈值的部位图片添加至车辆部位损伤识别模型的训练样本中重新训练车辆部位损伤识别模型，使车辆部位损伤识别模型得到强化学习，形成模型的自适应效果，通过不断强化和更新，提高车辆部位损伤识别模型模型识别准确率。
如图2所示，是本申请车辆定损方法的第二较佳实施例的流程图。根据不同的需求，该流程图中步骤的顺序可以改变，某些步骤可以省略。
S20至S24分别与第一较佳实施例中的S10至S14对应，在此不再详述。
S25、当车辆为投保车辆时，所述电子设备将所述车辆图片输入至训练好的车型识别模型中，输出车辆的品牌和车型。优选实施方式，将所述车辆图片的全景图片作为所述车型识别模型的输入。在训练车型识别模型时，车型识别模型的训练样本为各种品牌车辆的各种车型的全景图片。当车型识别模型训练好后，就能自动识别输入的全景图片中车辆的品牌和车型，训练算法为现有技术，包括，但不限于：卷积神经网络模型。
可选地，在所述车型识别模型的训练过程包括：
A、配置各种品牌各种车型(例如，奥迪Q5的全景图片、奥迪A3的全景图片、奔驰C级的全景图片、奔驰E级的全景图片等)对应的预设数量的样本图片；
B、将各个样本图片进行图片预处理以获得训练所述车型识别模型模型训练图片，例如可通过对各个样本图片进行图片预处理如缩放、裁剪、翻转及/或扭曲等操作后，使训练图片具有相同的尺寸及相同的视角后，才进行模型训练，以有效提高模型训练的真实性及准确率。
C、将所有训练图片分为第一比例(例如，85％)的训练集、第二比例(例如，15％)的验证集；
D、利用所述训练集训练所述车型识别模型模型；
E、利用所述验证集验证训练的车型识别模型模型的准确率，若准确率大于或者等于预设准确率，则训练结束，或者，若准确率小于预设准确率，则增加各个部位对应的样本图片；
数量并重新执行上述步骤B、C、D、E，直至训练的车型识别模型的准确率大于或者等于预设准确率。
S26、所述电子设备根据车辆的品牌和车型,及车辆的各个部位的最终定损结果，确定车辆的各个部位的维修数据。
在可选实施例中，所述维修数据包括维修配件的价格数据、维修工时费用数据。各个部位的不同的损伤程度对应不同的维修配件及不同的维修工时。从数据库中读取该品牌和车型的车辆的各个部位的维修数据。当所述数据库没有该品牌和车型的车辆的各个部位的维修数据时，向供应商发送询问指令以询问该品牌和车型的车辆的各个部位的维修数据。
S27、所述电子设备根据车辆的各个部位的维修数据，计算车辆的维修费用，并发送给用户的用户设备。
在可选实施例中，将车辆的各个部位的维修配件的价格数据、维修工时费用数据进行累加，作为车辆的维修费用。
在可选实施例中，所述方法还包括：获取车辆的投保数据，根据投保数据及维修费用，确定理赔数据，将理赔数据发送给该车辆的用户的设备以供 用户查看。
通过上述实施例中，本申请能准确对损坏车辆进行定损，并根据车辆的定损情况及投保数据，计算车辆的维修费用和理赔数据，提高了理赔效率及理赔的透明化。
如图3所示，本申请车辆定损装置的第一较佳实施例的程序模块图。所述车辆定损装置3包括，但不限于以下一个或者多个模块：获取模块30、分割模块31、识别模块32、训练模块33、确定模块34、发送模块35及提示模块36。本申请所称的单元是指一种能够被车辆定损装置3的处理器所执行并且能够完成固定功能的一系列计算机程序段，其存储在存储器中。关于各单元的功能将在后续的实施例中详述。
所述获取模块30获取车辆图片。
在可选实施例中，对事故现场中的车辆进行拍照，将拍摄的车辆图片发送至云端，所述车辆图片包括车辆全景图片、车辆部分图片等等。
优选地，所述获取模块30还用于：对获取的车辆图片进行检测，并判断获取的车辆图片是否合格，当所述获取的车辆图片不合格时，提示用户重新上传车辆图片。对所述获取的图片的检测内容包含图片清晰度、拍摄角度、拍摄部位的可识别程度、图片是否存在篡改嫌疑等，例如图片清晰度是否在配置的清晰度范围内，拍摄角度是否在配置的角度范围内、拍摄部位的可识别程度是否在配置的可识别程度范围内等等。若获取的车辆图片合格时，执行分割模块31。这样可以避免不合格的图片对后续的定损结果造成的影响，以提高车辆定损的准确度。
所述分割模块31将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片。
在可选实施中，所述车辆部位分割模型用于分割车辆的各个部位图片。所述车辆部位分割模型的训练样本包括车辆的各个部位的图片，如门把手，车门，轮胎，等等。在训练车辆部位分割模型的过程中，车辆部位分割模型不断学习车辆的各个部位的特征。当车辆部位分割模型训练好后，可以从车辆图片中识别，并分割出车辆的各个部位图片，便于后续判断各个部位图片的损伤程度。
优选地，所述分割模块31还用于：从车辆图片中识别出车牌部位、VIN码部位，从车牌部位中识别出车牌号，从VIN码部位识别出VIN码，利用车牌号或者VIN码识别车辆是否为投保车辆，当为投保车辆时，判断车辆的各个部位图片的损伤程度。
可选地，训练模块33在所述车辆部位分割模型的训练过程包括：
A、配置各个部位(例如，门把手，车门，轮胎、左前门、右前门、左叶 子板、右叶子板、前保险杠、后保险杠等)对应的预设数量的样本图片；
B、将各个样本图片进行图片预处理以获得训练所述车辆部位分割模型训练图片，例如可通过对各个样本图片进行图片预处理如缩放、裁剪、翻转及/或扭曲等操作后，使训练图片具有相同的尺寸及相同的视角后，才进行模型训练，以有效提高模型训练的真实性及准确率。
C、将所有训练图片分为第一比例(例如，70％)的训练集、第二比例(例如，30％)的验证集；
D、利用所述训练集训练所述车辆部位分割模型；
E、利用所述验证集验证训练的车辆部位分割模型的准确率，若准确率大于或者等于预设准确率，则训练结束，或者，若准确率小于预设准确率，则增加各个部位对应的样本图片；
数量并重新执行上述步骤B、C、D、E，直至训练的车辆部位分割模型的准确率大于或者等于预设准确率。
识别模块32所述电子设备将各个部位图片输入至训练好的车辆部位损伤识别模型中，识别各个部位的机器定损结果及输出各个部位的机器定损结果的置信度。
在可选实施例中，所述车辆部位损伤识别模型用于判断各个部位的损伤程度，并输出各个部位的机器定损结果的置信度。训练车辆部位损伤识别模型的训练样本包括各个部位的各种损伤程度的图片。在训练车辆部位损伤识别模型的过程中，车辆部位损伤识别模型不断学习各个部位的各种损伤程度的特征。当车辆部位损伤识别模型训练好后，能对各个部位图片进行定损，判定各个部位的损伤程度，并输出各个部位的定损结果的置信度。后续可以将置信度低于阈值的机器定损结果发给多个专业人员进行定损。
优选地，所述确定模块34用于：将置信度高于阈值的部位的机器定损结果作为置信度高于阈值的部位的最终定损结果。
所述发送模块35将置信度低于或等于阈值的部位图片发送至定损人员的用户设备上以使定损人员对该部位图片进行定损，并确定置信度低于阈值的部位的最终定损结果。
在可选实施例中，定损人员为一个或者多个。将超过预设人数的定损结果作为人工定损结果。例如，将置信度低于阈值的部位图片发送至5个定损人员，有4个定损人员判定为一级损伤程度，1个定损人员判定为二级损伤程度，则人工定损结果为一级损伤程度。这样将置信度低于阈值的部位图片发送给多个定损人员以使定损人员对车辆定损，并采用大部分定损人员的相同的定损结果作为人工定损结果，可以有效避免人工干扰因素。
优选地，所述提示模块36用于：当将置信度低于阈值的部位的发送至定损人员时，提示用户等待定损结果。
优选地，所述确定模块34用于：当该部位的机器定损结果与人工定损结果不同时，将人工定损结果作为置信度低于阈值的部位的最终定损结果。这样可以人工干预车辆的定损结果，提高定损精度。
所述训练模块33将置信度低于阈值的部位图片添加至车辆部位损伤识别模型的训练样本中重新训练车辆部位损伤识别模型。
在可选实施例中，所述训练模块33将该置信度低于阈值的部位图片更新至车辆部位损伤识别模型的训练样本中。例如，该部位图片的机器定损结果为二级损伤，人工定损结果为一级损伤，则车辆部位损伤识别模型中二级损伤的模型参数需要强化，需要学习更多的该部位的二级损伤的特征，将置信度低于阈值的部位图片添加至二级损伤类别中，作为二级损伤类别的训练样本。这样可以增加机器学习算法判定错误的类别的样本，并对车辆部位损伤识别模型重新训练，使车辆部位损伤识别模型学习判定错误的类别的样本的特征，从而使车辆部位损伤识别模型的模型参数后续能准确的判定该判断错误的类别的车辆损伤程度。
可选地，所述训练模块33在所述车辆部位损伤识别模型的训练过程包括：
A、配置各个部位的各种损伤程度(例如，对于门把手部位，分别配置一级损伤程度、二级损伤程度、三级损伤程度等等各种损伤程度)对应的预设数量的样本图片；
B、将各个样本图片进行图片预处理以获得训练所述车辆部位损伤识别模型的训练图片，例如可通过对各个样本图片进行图片预处理如缩放、裁剪、翻转及/或扭曲等操作后，使训练图片具有相同的尺寸及相同的视角后，才进行模型训练，以有效提高模型训练的真实性及准确率。
C、将所有训练图片分为第一比例(例如，80％)的训练集、第二比例(例如，20％)的验证集；
D、利用所述训练集训练所述车辆部位损伤识别模型；
E、利用所述验证集验证训练的车辆部位损伤识别模型的准确率，若准确率大于或者等于预设准确率，则训练结束，或者，若准确率小于预设准确率，则增加各个部位对应的样本图片；
数量并重新执行上述步骤B、C、D、E，直至训练的车辆部位损伤识别模型的准确率大于或者等于预设准确率。
在可选实施例中，所述车辆部位分割模型、所述车辆部位损伤识别模型及所述车型识别模型可以是为不带有全连接层的深度卷积神经网络模型，所述深度卷积神经网络模型包括输入层、卷积层、池化层、上采样层及裁切层，在一种具体的实施方式中，所述深度卷积神经网络模型由1个输入层，16个卷积层，5个池化层，1个上采样层，1个裁切层构成。
由于在传统的分类问题中，一般需要用全连接层来输出每一张图片属于每个类的概率，然而在语义分割问题上，用这种方法来预测每个样本属于哪个类必然会导致效率低下。因此，本实施例中的识别模型为不带有全连接层的深度卷积神经网络模型，该深度卷积神经网络模型只需在Conv8上，用一个卷积层来输出每个像素的分类score。在该层上，特征图上的每个点都有class num+1个分类中不同分类的score，因此输出的channel也是class num+1，识别效率大大提高。
通过上述实施，本申请能将置信度低于阈值的部位图片添加至车辆部位损伤识别模型的训练样本中重新训练车辆部位损伤识别模型，使车辆部位损伤识别模型得到强化学习，形成模型的自适应效果，通过不断强化和更新，提高车辆部位损伤识别模型模型识别准确率。
如图4所示，本申请车辆定损装置的第二较佳实施例的程序模块图。所述车辆定损装置3除了包括第一较佳实施中的一个或者多个模块：获取模块30、分割模块31、识别模块32、训练模块33、确定模块34、发送模块35及提示模块36之外，所述车辆定损装置3还可以包括以下一个或者多个模块：输出模块37及计算模块38。本申请所称的单元是指一种能够被车辆定损装置3的处理器所执行并且能够完成固定功能的一系列计算机程序段，其存储在存储器中。关于各单元的功能将在后续的实施例中详述。
当车辆为投保车辆时，所述输出模块37将所述车辆图片输入至训练好的车型识别模型中，输出车辆的品牌和车型。优选实施方式，将所述车辆图片的全景图片作为所述车型识别模型的输入。在训练车型识别模型时，车型识别模型的训练样本为各种品牌车辆的各种车型的全景图片。当车型识别模型训练好后，就能自动识别输入的全景图片中车辆的品牌和车型，训练算法为现有技术，包括，但不限于：卷积神经网络模型。
可选地，所述训练模块33在所述车型识别模型的训练过程包括：
A、配置各种品牌各种车型(例如，奥迪Q5的全景图片、奥迪A3的全景图片、奔驰C级的全景图片、奔驰E级的全景图片等)对应的预设数量的样本图片；
B、将各个样本图片进行图片预处理以获得训练所述车型识别模型模型训练图片，例如可通过对各个样本图片进行图片预处理如缩放、裁剪、翻转及/或扭曲等操作后，使训练图片具有相同的尺寸及相同的视角后，才进行模型训练，以有效提高模型训练的真实性及准确率。
C、将所有训练图片分为第一比例(例如，85％)的训练集、第二比例(例如，15％)的验证集；
D、利用所述训练集训练所述车型识别模型模型；
E、利用所述验证集验证训练的车型识别模型模型的准确率，若准确率大于或者等于预设准确率，则训练结束，或者，若准确率小于预设准确率，则增加各个部位对应的样本图片；
数量并重新执行上述步骤B、C、D、E，直至训练的车型识别模型的准确率大于或者等于预设准确率。
所述确定模块34根据车辆的品牌和车型，及车辆的各个部位的最终定损结果，确定车辆的各个部位的维修数据。
在可选实施例中，所述维修数据包括维修配件的价格数据、维修工时费用数据。各个部位的不同的损伤程度对应不同的维修配件及不同的维修工时。从数据库中读取该品牌和车型的车辆的各个部位的维修数据。当所述数据库没有该品牌和车型的车辆的各个部位的维修数据时，向供应商发送询问指令以询问该品牌和车型的车辆的各个部位的维修数据。
所述计算模块38根据车辆的各个部位的维修数据，计算车辆的维修费用，并发送给用户的用户设备。
在可选实施例中，将车辆的各个部位的维修配件的价格数据、维修工时费用数据进行累加，作为车辆的维修费用。
在可选实施例中，所述确定模块34还用于：获取车辆的投保数据，根据投保数据及维修费用，确定理赔数据，将理赔数据发送给该车辆的用户的设备以供用户查看。
通过上述实施例中，本申请能准确对损坏车辆进行定损，并根据车辆的定损情况及投保数据，计算车辆的维修费用和理赔数据，提高了理赔效率及理赔的透明化。
上述以软件程序模块的形式实现的集成的单元，可以存储在一个计算机可读取存储介质中。上述软件程序模块存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)或处理器(processor)执行本申请每个实施例所述方法的部分步骤。
如图5所示，所述电子设备5包括至少一个发送装置51、至少一个存储器52、至少一个处理器53、至少一个接收装置54以及至少一个通信总线。其中，所述通信总线用于实现这些组件之间的连接通信。
所述电子设备5是一种能够按照事先设定或存储的指令，自动进行数值计算和/或信息处理的设备，其硬件包括但不限于微处理器、专用集成电路(Application Specific Integrated Circuit，ASIC)、可编程门阵列(Field－Programmable Gate Array，FPGA)、数字处理器(Digital Signal Processor，DSP)、嵌入式设备等。所述电子设备5还可包括网络设备和/或用户设备。其中，所述网络设备包括但不限于单个网络服务器、多个网络服务器组成的服务器组 或基于云计算(Cloud Computing)的由大量主机或网络服务器构成的云，其中，云计算是分布式计算的一种，由一群松散耦合的计算机集组成的一个超级虚拟计算机。
所述电子设备5可以是，但不限于任何一种可与用户通过键盘、触摸板或声控设备等方式进行人机交互的电子产品，例如，平板电脑、智能手机、个人数字助理(Personal Digital Assistant，PDA)、智能式穿戴式设备、摄像设备、监控设备等终端。
所述电子设备5所处的网络包括，但不限于互联网、广域网、城域网、局域网、虚拟专用网络(Virtual Private Network，VPN)等。
其中，所述接收装置54和所述发送装置51可以是有线发送端口，也可以为无线设备，例如包括天线装置，用于与其他设备进行数据通信。
所述存储器52用于存储程序代码。所述存储器52可以是集成电路中没有实物形式的具有存储功能的电路，如RAM(Random-Access Memory，随机存取存储器)、FIFO(First In First Out，)等。或者，所述存储器52也可以是具有实物形式的存储器，如内存条、TF卡(Trans-flash Card)、智能媒体卡(smart media card)、安全数字卡(secure digital card)、快闪存储器卡(flash card)等储存设备等等。
所述处理器53可以包括一个或者多个微处理器、数字处理器。所述处理器53可调用存储器52中存储的程序代码以执行相关的功能。例如，图3及图4中所述的各个模块是存储在所述存储器52中的程序代码，并由所述处理器53所执行，以实现一种车辆定损方法。所述处理器53又称中央处理器(CPU，Central Processing Unit)，是一块超大规模的集成电路，是运算核心(Core)和控制核心(Control Unit)。
本申请实施例还提供一种计算机可读存储介质，其上存储有计算机指令，所述指令当被包括一个或多个处理器的电子设备执行时，使电子设备执行如上文方法实施例所述的车辆定损方法。
结合图1及图2所示，所述电子设备5中的所述存储器52存储多个指令以实现一种车辆定损方法，所述处理器53可执行所述多个指令从而实现：
获取车辆图片；将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片；将各个部位图片输入至训练好的车辆部位损伤识别模型中，识别各个部位的机器定损结果及输出各个部位的机器定损结果的置信度；将置信度低于或等于阈值的部位图片发送至定损人员的用户设备上以使定损人员对该部位图片进行定损，并确定置信度低于阈值的部位的最终定损结果；将置信度低于阈值的部位图片添加至车辆部位损伤识别模型的训练样本中，重新训练车辆部位损伤识别模型。
所述处理器执行所述多个指令时还执行以下指令：在将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片之前，对获取的车辆图片进行检测，判断获取的车辆图片是否合格，检测的内容包括以下一种或者多种的组合：图片清晰度、拍摄角度、拍摄部位的可识别程度、图片是否存在篡改嫌疑；
当所述获取的车辆图片不合格时，提示用户重新上传车辆图片。
所述处理器执行所述多个指令时还执行以下指令：在将各个部位图片输入至训练好的车辆部位损伤识别模型之前，从车辆图片中识别出车牌部位、VIN码部位，从车牌部位中识别出车牌号，从VIN码部位识别出VIN码，利用车牌号或者VIN码识别车辆是否为投保车辆，当为投保车辆时，判断车辆的各个部位图片的损伤程度。
所述处理器执行所述多个指令时还执行以下指令：当将置信度低于阈值的部位的发送至定损人员时，提示用户等待定损结果。
所述处理器执行所述多个指令时还执行以下指令：当该部位的机器定损结果与人工定损结果不同时，将人工定损结果作为置信度低于阈值的部位的最终定损结果。
所述处理器执行所述多个指令时还执行以下指令：将置信度高于阈值的部位的机器定损结果作为置信度高于阈值的部位的最终定损结果。
所述处理器执行所述多个指令时还执行以下指令：
当车辆为投保车辆时，将所述车辆图片输入至训练好的车型识别模型中，输出车辆的品牌和车型；
根据车辆的品牌和车型,及车辆的各个部位的最终定损结果，确定车辆的各个部位的维修数据；
根据车辆的各个部位的维修数据，计算车辆的维修费用，并发送给用户的用户设备。
所述处理器执行所述多个指令时还执行以下指令：获取车辆的投保数据，根据投保数据及维修费用，确定理赔数据，将理赔数据发送给该车辆的用户的设备以供用户查看。
在任意实施例中所述车辆定损方法对应的多个指令存储在所述存储器52，并通过所述处理器53来执行，在此不再详述。
以上说明的本申请的特征性的手段可以通过集成电路来实现，并控制实现上述任意实施例中所述车辆定损方法的功能。即，本申请的集成电路安装于所述电子设备中，使所述电子设备发挥如下功能：获取车辆图片；将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片；将各个部位图片输入至训练好的车辆部位损伤识别模型中，识别各个部位的机器定损结果及输出 各个部位的机器定损结果的置信度；将置信度低于或等于阈值的部位图片发送至定损人员的用户设备上以使定损人员对该部位图片进行定损，并确定置信度低于阈值的部位的最终定损结果；将置信度低于阈值的部位图片添加至车辆部位损伤识别模型的训练样本中，重新训练车辆部位损伤识别模型。
在任意实施例中所述车辆定损方法所能实现的功能都能通过本申请的集成电路安装于所述电子设备中，使所述电子设备发挥任意实施例中所述车辆定损方法所能实现的功能，在此不再详述。
需要说明的是，对于前述的各方法实施例，为了简单描述，故将其都表述为一系列的动作组合，但是本领域技术人员应该知悉，本申请并不受所描述的动作顺序的限制，因为依据本申请，某些步骤可以采用其他顺序或者同时进行。其次，本领域技术人员也应该知悉，说明书中所描述的实施例均属于优选实施例，所涉及的动作和模块并不一定是本申请所必须的。
在上述实施例中，对各个实施例的描述都各有侧重，某个实施例中没有详述的部分，可以参见其他实施例的相关描述。
在本申请所提供的几个实施例中，应该理解到，所揭露的装置，可通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如所述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请的各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
所述集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。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或者该技术方案的全部或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一个存储介质中，包括若干指令用以使得一台计算机设备(可为个人计算机、服务器或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。 而前述的存储介质包括：U盘、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、移动硬盘、磁碟或者光盘等各种可以存储程序代码的介质。
以上所述，以上实施例仅用以说明本申请的技术方案，而非对其限制；尽管参照前述实施例对本申请进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本申请各实施例技术方案的范围。

Claims (20)

  1. 一种车辆定损方法，其特征在于，所述方法包括：
    获取车辆图片；
    将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片；
    将各个部位图片输入至训练好的车辆部位损伤识别模型中，识别各个部位的机器定损结果及输出各个部位的机器定损结果的置信度；
    将置信度低于或等于阈值的部位图片发送至定损人员的用户设备上以使定损人员对该部位图片进行定损，并确定置信度低于阈值的部位的最终定损结果；
    将置信度低于阈值的部位图片添加至车辆部位损伤识别模型的训练样本中，重新训练车辆部位损伤识别模型。
  2. 如权利要求1所述的车辆定损方法，其特征在于，在将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片之前，所述方法还包括：
    对获取的车辆图片进行检测，判断获取的车辆图片是否合格，检测的内容包括以下一种或者多种的组合：图片清晰度、拍摄角度、拍摄部位的可识别程度、图片是否存在篡改嫌疑；
    当所述获取的车辆图片不合格时，提示用户重新上传车辆图片。
  3. 如权利要求1所述的车辆定损方法，其特征在于，在将各个部位图片输入至训练好的车辆部位损伤识别模型之前，所述方法还包括：
    从车辆图片中识别出车牌部位、VIN码部位，从车牌部位中识别出车牌号，从VIN码部位识别出VIN码，利用车牌号或者VIN码识别车辆是否为投保车辆，当为投保车辆时，判断车辆的各个部位图片的损伤程度。
  4. 如权利要求1所述的车辆定损方法，其特征在于，所述方法还包括：当该部位的机器定损结果与人工定损结果不同时，将人工定损结果作为置信度低于阈值的部位的最终定损结果。
  5. 如权利要求1所述的车辆定损方法，其特征在于，所述方法还包括：将置信度高于阈值的部位的机器定损结果作为置信度高于阈值的部位的最终定损结果。
  6. 如权利要求1所述的车辆定损方法，其特征在于，所述方法还包括：
    当车辆为投保车辆时，将所述车辆图片输入至训练好的车型识别模型中，输出车辆的品牌和车型；
    根据车辆的品牌和车型，及车辆的各个部位的最终定损结果，确定车辆的各个部位的维修数据；
    根据车辆的各个部位的维修数据，计算车辆的维修费用，并发送给用户的用户设备。
  7. 如权利要求6所述的车辆定损方法，其特征在于，所述方法还包括：获 取车辆的投保数据，根据投保数据及维修费用，确定理赔数据，将理赔数据发送给该车辆的用户的设备以供用户查看。
  8. 一种车辆定损装置，所述装置包括：
    获取模块，用于获取车辆图片；
    分割模块，用于将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片；
    识别模块，用于将各个部位图片输入至训练好的车辆部位损伤识别模型中，识别各个部位的机器定损结果及输出各个部位的机器定损结果的置信度；
    发送模块，用于将置信度低于或等于阈值的部位图片发送至定损人员的用户设备上以使定损人员对该部位图片进行定损，并确定置信度低于阈值的部位的最终定损结果；
    训练模块，用于将置信度低于阈值的部位图片添加至车辆部位损伤识别模型的训练样本中，重新训练车辆部位损伤识别模型。
  9. 一种电子设备，其特征在于，所述电子设备包括存储器及处理器，所述存储器用于存储至少一个计算机可读指令，所述处理器用于执行所述至少一个计算机可读指令以实现以下步骤：
    获取车辆图片；
    将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片；
    将各个部位图片输入至训练好的车辆部位损伤识别模型中，识别各个部位的机器定损结果及输出各个部位的机器定损结果的置信度；
    将置信度低于或等于阈值的部位图片发送至定损人员的用户设备上以使定损人员对该部位图片进行定损，并确定置信度低于阈值的部位的最终定损结果；
    将置信度低于阈值的部位图片添加至车辆部位损伤识别模型的训练样本中，重新训练车辆部位损伤识别模型。
  10. 如权利要求9所述的电子设备，其特征在于，在将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片之前，所述处理器还用于执行所述至少一个计算机可读指令，以实现以下步骤：
    对获取的车辆图片进行检测，判断获取的车辆图片是否合格，检测的内容包括以下一种或者多种的组合：图片清晰度、拍摄角度、拍摄部位的可识别程度、图片是否存在篡改嫌疑；
    当所述获取的车辆图片不合格时，提示用户重新上传车辆图片。
  11. 如权利要求9所述的电子设备，其特征在于，在将各个部位图片输入至训练好的车辆部位损伤识别模型之前，所述处理器还用于执行所述至少一个计算机可读指令，以实现以下步骤：
    从车辆图片中识别出车牌部位、VIN码部位，从车牌部位中识别出车牌号，从VIN码部位识别出VIN码，利用车牌号或者VIN码识别车辆是否为投保车辆，当为投保车辆时，判断车辆的各个部位图片的损伤程度。
  12. 如权利要求9所述的电子设备，其特征在于，所述处理器还用于执行所 述至少一个计算机可读指令，以实现以下步骤：当该部位的机器定损结果与人工定损结果不同时，将人工定损结果作为置信度低于阈值的部位的最终定损结果。
  13. 如权利要求9所述的电子设备，其特征在于，所述处理器还用于执行所述至少一个计算机可读指令，以实现以下步骤：将置信度高于阈值的部位的机器定损结果作为置信度高于阈值的部位的最终定损结果。
  14. 如权利要求9所述的电子设备，其特征在于，所述处理器还用于执行所述至少一个计算机可读指令，以实现以下步骤：
    当车辆为投保车辆时，将所述车辆图片输入至训练好的车型识别模型中，输出车辆的品牌和车型；
    根据车辆的品牌和车型，及车辆的各个部位的最终定损结果，确定车辆的各个部位的维修数据；
    根据车辆的各个部位的维修数据，计算车辆的维修费用，并发送给用户的用户设备。
  15. 如权利要求14所述的电子设备，其特征在于，所述处理器还用于执行所述至少一个计算机可读指令，以实现以下步骤：获取车辆的投保数据，根据投保数据及维修费用，确定理赔数据，将理赔数据发送给该车辆的用户的设备以供用户查看。
  16. 一种非易失性可读存储介质，其特征在于，所述计算机可读存储介质存储有至少一个计算机可读指令，所述至少一个计算机可读指令被处理器执行时实现以下步骤：
    获取车辆图片；
    将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片；
    将各个部位图片输入至训练好的车辆部位损伤识别模型中，识别各个部位的机器定损结果及输出各个部位的机器定损结果的置信度；
    将置信度低于或等于阈值的部位图片发送至定损人员的用户设备上以使定损人员对该部位图片进行定损，并确定置信度低于阈值的部位的最终定损结果；
    将置信度低于阈值的部位图片添加至车辆部位损伤识别模型的训练样本中，重新训练车辆部位损伤识别模型。
  17. 如权利要求16所述的非易失性可读存储介质，其特征在于，在将车辆图片输入至训练好的车辆部位分割模型中，分割出车辆的各个部位图片之前，所述至少一个计算机可读指令被处理器执行时，还实现以下步骤：
    对获取的车辆图片进行检测，判断获取的车辆图片是否合格，检测的内容包括以下一种或者多种的组合：图片清晰度、拍摄角度、拍摄部位的可识别程度、图片是否存在篡改嫌疑；
    当所述获取的车辆图片不合格时，提示用户重新上传车辆图片。
  18. 如权利要求16所述的非易失性可读存储介质，其特征在于，在将各个部位图片输入至训练好的车辆部位损伤识别模型之前，所述至少一个计算机可读指令被处理器执行时，还实现以下步骤：
    从车辆图片中识别出车牌部位、VIN码部位，从车牌部位中识别出车牌号，从VIN码部位识别出VIN码，利用车牌号或者VIN码识别车辆是否为投保车辆，当为投保车辆时，判断车辆的各个部位图片的损伤程度。
  19. 如权利要求16所述的非易失性可读存储介质，其特征在于，所述至少一个计算机可读指令被处理器执行时，还实现以下步骤：当该部位的机器定损结果与人工定损结果不同时，将人工定损结果作为置信度低于阈值的部位的最终定损结果。
  20. 如权利要求16所述的非易失性可读存储介质，其特征在于，所述至少一个计算机可读指令被处理器执行时，还实现以下步骤：将置信度高于阈值的部位的机器定损结果作为置信度高于阈值的部位的最终定损结果。
PCT/CN2018/082577 2018-03-09 2018-04-10 车辆定损方法、装置、电子设备及存储介质 WO2019169688A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810196561.1 2018-03-09
CN201810196561.1A CN108446618A (zh) 2018-03-09 2018-03-09 车辆定损方法、装置、电子设备及存储介质

Publications (1)

Publication Number Publication Date
WO2019169688A1 true WO2019169688A1 (zh) 2019-09-12

Family

ID=63194477

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/082577 WO2019169688A1 (zh) 2018-03-09 2018-04-10 车辆定损方法、装置、电子设备及存储介质

Country Status (2)

Country Link
CN (1) CN108446618A (zh)
WO (1) WO2019169688A1 (zh)

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110889585A (zh) * 2019-10-12 2020-03-17 中国平安财产保险股份有限公司 信息分类决策方法、装置、计算机系统及可读存储介质
CN111415018A (zh) * 2020-03-18 2020-07-14 上海钧正网络科技有限公司 一种维修出库的评定方法、装置、介质及设备
CN111666973A (zh) * 2020-04-29 2020-09-15 平安科技（深圳）有限公司 车辆损伤图片处理方法、装置、计算机设备及存储介质
CN111915446A (zh) * 2020-08-14 2020-11-10 南京三百云信息科技有限公司 事故车辆定损方法、装置及终端设备
CN112085610A (zh) * 2020-09-07 2020-12-15 中国平安财产保险股份有限公司 目标物定损方法、装置、电子设备及计算机可读存储介质
CN112712498A (zh) * 2020-12-25 2021-04-27 北京百度网讯科技有限公司 移动终端执行的车辆定损方法、装置、移动终端、介质
CN113326954A (zh) * 2021-06-25 2021-08-31 中国平安财产保险股份有限公司 车辆维修任务调度方法、装置、设备及存储介质
CN113627252A (zh) * 2021-07-07 2021-11-09 浙江吉利控股集团有限公司 一种车辆定损方法、装置、存储介质及电子设备
CN113744392A (zh) * 2021-08-30 2021-12-03 深圳壹账通智能科技有限公司 三维模型库构建方法、装置、设备及介质
CN114140430A (zh) * 2021-11-30 2022-03-04 北京比特易湃信息技术有限公司 一种基于深度学习的车辆报损方法
CN114273245A (zh) * 2021-11-10 2022-04-05 上海艾豚科技有限公司 一种汽车内饰件多品种工件连续混合检测方法及系统

Families Citing this family (36)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110570317B (zh) * 2018-08-31 2023-06-20 创新先进技术有限公司 用于车辆核损的方法及装置
CN109271908B (zh) * 2018-09-03 2022-05-13 创新先进技术有限公司 车损检测方法、装置及设备
CN110570435B (zh) * 2018-09-10 2020-06-26 阿里巴巴集团控股有限公司 用于对车辆损伤图像进行损伤分割的方法及装置
CN109410270B (zh) * 2018-09-28 2020-10-27 百度在线网络技术（北京）有限公司 一种定损方法、设备和存储介质
CN109523556A (zh) * 2018-09-30 2019-03-26 百度在线网络技术（北京）有限公司 车辆部件分割方法和装置
CN109359676A (zh) * 2018-10-08 2019-02-19 百度在线网络技术（北京）有限公司 用于生成车辆损伤信息的方法和装置
CN109410218B (zh) * 2018-10-08 2020-08-11 百度在线网络技术（北京）有限公司 用于生成车辆损伤信息的方法和装置
CN109389169A (zh) * 2018-10-08 2019-02-26 百度在线网络技术（北京）有限公司 用于处理图像的方法和装置
CN109215027B (zh) * 2018-10-11 2024-05-24 平安科技（深圳）有限公司 一种基于神经网络的车辆定损方法、服务器及介质
CN109710654A (zh) * 2018-10-17 2019-05-03 青岛腾信汽车网络科技服务有限公司 一种车辆碰撞受损等级鉴定方法
CN109767339A (zh) * 2018-12-03 2019-05-17 中国人民财产保险股份有限公司 一种事故车辆的理赔数据确定方法、装置及系统
CN109614935B (zh) * 2018-12-12 2021-07-06 泰康保险集团股份有限公司 车辆定损方法及装置、存储介质及电子设备
CN109344819A (zh) * 2018-12-13 2019-02-15 深源恒际科技有限公司 基于深度学习的车辆损伤识别方法
CN109784171A (zh) * 2018-12-14 2019-05-21 平安科技（深圳）有限公司 车辆定损图像筛选方法、装置、可读存储介质及服务器
CN110569701B (zh) * 2018-12-29 2020-08-07 阿里巴巴集团控股有限公司 计算机执行的车辆定损方法及装置
CN110009509B (zh) * 2019-01-02 2021-02-19 创新先进技术有限公司 评估车损识别模型的方法及装置
CN109815997A (zh) * 2019-01-04 2019-05-28 平安科技（深圳）有限公司 基于深度学习的识别车辆损伤的方法和相关装置
CN109740547A (zh) * 2019-01-04 2019-05-10 平安科技（深圳）有限公司 一种图像处理方法、设备及计算机可读存储介质
CN110135437B (zh) * 2019-05-06 2022-04-05 北京百度网讯科技有限公司 用于车辆的定损方法、装置、电子设备和计算机存储介质
CN110502998B (zh) * 2019-07-23 2023-01-31 平安科技（深圳）有限公司 车辆定损方法、装置、设备和存储介质
CN110660000A (zh) * 2019-09-09 2020-01-07 平安科技（深圳）有限公司 数据预测方法、装置、设备及计算机可读存储介质
CN110837779A (zh) * 2019-10-12 2020-02-25 平安科技（深圳）有限公司 车辆外观智能诊断方法、装置及计算机可读存储介质
CN111191400B (zh) * 2019-12-31 2023-12-29 上海钧正网络科技有限公司 基于用户报障数据的车辆零部件寿命预测方法及系统
CN111242070A (zh) * 2020-01-19 2020-06-05 上海眼控科技股份有限公司 目标物体检测方法、计算机设备和存储介质
CN111489433B (zh) * 2020-02-13 2023-04-25 北京百度网讯科技有限公司 车辆损伤定位的方法、装置、电子设备以及可读存储介质
CN113496242A (zh) * 2020-04-07 2021-10-12 华晨宝马汽车有限公司 对车辆的损伤部位进行分类的方法和设备
CN111583215B (zh) * 2020-04-30 2024-07-02 平安科技（深圳）有限公司 损伤图像智能定损方法、装置、电子设备及存储介质
CN111666990A (zh) * 2020-05-27 2020-09-15 平安科技（深圳）有限公司 车辆损伤特征检测方法、装置、计算机设备及存储介质
CN111709352B (zh) * 2020-06-12 2022-10-04 浪潮集团有限公司 一种基于神经网络的车辆划痕检测方法
CN112017065B (zh) * 2020-08-27 2024-05-24 中国平安财产保险股份有限公司 车辆定损理赔方法、装置及计算机可读存储介质
CN112085721A (zh) * 2020-09-07 2020-12-15 中国平安财产保险股份有限公司 基于人工智能的水淹车定损方法、装置、设备及存储介质
CN112700436A (zh) * 2021-01-13 2021-04-23 上海微亿智造科技有限公司 一种用于提高工业质检模型迭代的方法、系统及介质
CN113378619B (zh) * 2021-03-12 2023-07-04 中国平安财产保险股份有限公司 保险业务数据处理方法、装置、电子设备和存储介质
CN113436175B (zh) * 2021-06-30 2023-08-18 平安科技（深圳）有限公司 车图像分割质量的评估方法、装置、设备及存储介质
CN113947690A (zh) * 2021-10-14 2022-01-18 中国平安财产保险股份有限公司 业务清单数据的生成方法、装置、设备及存储介质
CN116168356B (zh) * 2023-04-26 2023-07-21 威海海洋职业学院 一种基于计算机视觉的车辆损伤判别方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105488789A (zh) * 2015-11-24 2016-04-13 大连楼兰科技股份有限公司 汽车零部件分级定损方法
US20170316262A1 (en) * 2016-04-30 2017-11-02 Infrared Integrated Systems Limited System, Method And Apparatus For Occupancy Detection
CN107358596A (zh) * 2017-04-11 2017-11-17 阿里巴巴集团控股有限公司 一种基于图像的车辆定损方法、装置、电子设备及系统
CN107403424A (zh) * 2017-04-11 2017-11-28 阿里巴巴集团控股有限公司 一种基于图像的车辆定损方法、装置及电子设备

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106231257A (zh) * 2016-07-29 2016-12-14 深圳市永兴元科技有限公司 车辆定损方法及车辆定损服务器
CN106303426A (zh) * 2016-08-17 2017-01-04 苏州华兴源创电子科技有限公司 一种车辆远程定损方法及系统
CN107133876A (zh) * 2017-05-08 2017-09-05 明觉科技（北京）有限公司 车辆定损方法及定损客户端

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105488789A (zh) * 2015-11-24 2016-04-13 大连楼兰科技股份有限公司 汽车零部件分级定损方法
US20170316262A1 (en) * 2016-04-30 2017-11-02 Infrared Integrated Systems Limited System, Method And Apparatus For Occupancy Detection
CN107358596A (zh) * 2017-04-11 2017-11-17 阿里巴巴集团控股有限公司 一种基于图像的车辆定损方法、装置、电子设备及系统
CN107403424A (zh) * 2017-04-11 2017-11-28 阿里巴巴集团控股有限公司 一种基于图像的车辆定损方法、装置及电子设备

Cited By (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110889585A (zh) * 2019-10-12 2020-03-17 中国平安财产保险股份有限公司 信息分类决策方法、装置、计算机系统及可读存储介质
CN110889585B (zh) * 2019-10-12 2023-08-22 中国平安财产保险股份有限公司 信息分类决策方法、装置、计算机系统及可读存储介质
CN111415018A (zh) * 2020-03-18 2020-07-14 上海钧正网络科技有限公司 一种维修出库的评定方法、装置、介质及设备
CN111666973A (zh) * 2020-04-29 2020-09-15 平安科技（深圳）有限公司 车辆损伤图片处理方法、装置、计算机设备及存储介质
CN111666973B (zh) * 2020-04-29 2024-04-09 平安科技（深圳）有限公司 车辆损伤图片处理方法、装置、计算机设备及存储介质
CN111915446A (zh) * 2020-08-14 2020-11-10 南京三百云信息科技有限公司 事故车辆定损方法、装置及终端设备
CN112085610A (zh) * 2020-09-07 2020-12-15 中国平安财产保险股份有限公司 目标物定损方法、装置、电子设备及计算机可读存储介质
CN112085610B (zh) * 2020-09-07 2023-08-22 中国平安财产保险股份有限公司 目标物定损方法、装置、电子设备及计算机可读存储介质
CN112712498A (zh) * 2020-12-25 2021-04-27 北京百度网讯科技有限公司 移动终端执行的车辆定损方法、装置、移动终端、介质
EP3869404A3 (en) * 2020-12-25 2022-01-26 Beijing Baidu Netcom Science And Technology Co. Ltd. Vehicle loss assessment method executed by mobile terminal, device, mobile terminal and medium
CN113326954B (zh) * 2021-06-25 2023-07-07 中国平安财产保险股份有限公司 车辆维修任务调度方法、装置、设备及存储介质
CN113326954A (zh) * 2021-06-25 2021-08-31 中国平安财产保险股份有限公司 车辆维修任务调度方法、装置、设备及存储介质
CN113627252A (zh) * 2021-07-07 2021-11-09 浙江吉利控股集团有限公司 一种车辆定损方法、装置、存储介质及电子设备
CN113744392A (zh) * 2021-08-30 2021-12-03 深圳壹账通智能科技有限公司 三维模型库构建方法、装置、设备及介质
CN114273245A (zh) * 2021-11-10 2022-04-05 上海艾豚科技有限公司 一种汽车内饰件多品种工件连续混合检测方法及系统
CN114140430A (zh) * 2021-11-30 2022-03-04 北京比特易湃信息技术有限公司 一种基于深度学习的车辆报损方法

Also Published As

Publication number Publication date
CN108446618A (zh) 2018-08-24

Similar Documents

Publication Publication Date Title
WO2019169688A1 (zh) 车辆定损方法、装置、电子设备及存储介质
CN110020592B (zh) 物体检测模型训练方法、装置、计算机设备及存储介质
WO2021077984A1 (zh) 对象识别方法、装置、电子设备及可读存储介质
WO2019200781A1 (zh) 票据识别方法、装置及存储介质
WO2017220032A1 (zh) 基于深度学习的车牌分类方法、系统、电子装置及存储介质
WO2019120115A1 (zh) 人脸识别的方法、装置及计算机装置
US10152644B2 (en) Progressive vehicle searching method and device
WO2019237846A1 (zh) 图像处理方法、人脸识别方法、装置和计算机设备
WO2018166116A1 (zh) 车损识别方法、电子装置及计算机可读存储介质
WO2019127924A1 (zh) 样本权重分配方法、模型训练方法、电子设备及存储介质
CN106897746B (zh) 数据分类模型训练方法和装置
CN111340008A (zh) 对抗补丁生成、检测模型训练、对抗补丁防御方法及系统
CN107679997A (zh) 医疗理赔拒付方法、装置、终端设备及存储介质
CN110765860A (zh) 摔倒判定方法、装置、计算机设备及存储介质
CN107992807B (zh) 一种基于cnn模型的人脸识别方法及装置
WO2020047854A1 (en) Detecting objects in video frames using similarity detectors
CN111598182A (zh) 训练神经网络及图像识别的方法、装置、设备及介质
CN111680544B (zh) 人脸识别方法、装置、系统、设备及介质
CN107832794A (zh) 一种卷积神经网络生成方法、车系识别方法及计算设备
JP2011248879A5 (zh)
CN112712068B (zh) 一种关键点检测方法、装置、电子设备及存储介质
EP3859673A1 (en) Model generation
TWI803243B (zh) 圖像擴增方法、電腦設備及儲存介質
CN110175500B (zh) 指静脉比对方法、装置、计算机设备及存储介质
WO2021042544A1 (zh) 基于去网纹模型的人脸验证方法、装置、计算机设备及存储介质

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: 18908642

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 PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 10/12/2020)

122 Ep: pct application non-entry in european phase

Ref document number: 18908642

Country of ref document: EP

Kind code of ref document: A1