Vehicle damage image enhancement method and device

Abstract

The invention provides a vehicle damage image enhancement method and device. The invention provides a set of image restoration scheme based on the convolutional neural network, the image can be restored, a reflective area and a shadow area in the image can be eliminated, and the reflective and shadow area in the component area can be restored by segmenting each component area in the vehicle exterior part. The accuracy of a subsequent AI loss assessment link and an artificial loss checking link can be greatly improved, and loss can be retrieved for insurance companies.

Description

technical field [0001] The present application relates to the technical field of image data processing, in particular to a method and device for enhancing car damage images. Background technique [0002] When making auto insurance claims, vehicle damage determination refers to listing the repair man-hours, replacement items and amounts item by item for the vehicle loss parts and repair items caused by the accident according to the on-site survey records. There are two main types of vehicle damage assessment schemes currently available to insurance companies: [0003] The first is the manual damage assessment plan: the damage assessment specialist will go to the accident site to assess the damage of the vehicle based on experience and give a damage assessment sheet, and combine the 4S shop to give the repair man-hour cost, replacement items and amount; [0004] The second is the intelligent loss assessment solution: the loss assessment commissioner or the car owner uses the ...

AI Technical Summary

Problems solved by technology

However, when the AI ​​damage assessment system assesses the damage of the case image, the quality of the image is relatively high, and because the time of the traffic accident is uncertain, it may occur during the day or at night, so when collecting accident photos The lighting conditions are also complex, so there may be reflections, shadows, etc. in the collected vehicle damage images. These images often have poor accuracy when using the AI ​​​​damage assessment system for damage assessment. For example, the Reflective identification as scratches etc.

Images