A CNN-based Canny Edge Detection Approach for Car Scratch Detection

Image-based vehicle protection processing is a significant technological scope with a huge degree of computerization. In this research work, the advancement and modeling of a novel design proposal to identify scratches in vehicles using image processing and a deep learning is performed. This works also present and android application by implementing the proposed model for car scratch detection. As the image dataset included the images of the vehicle are captured in a different view for the image processing and detection, scratch in the vehicles can be detected using a canny edge detection scheme with different parts on the surface of the vehicle and it reduces the noise of the image. The convolutional neural network (CNN) is applied here to classify scratch zone and non-scratch zone in the surface area of the vehicle. The experimental result shows that the proposed system has increased detection accuracy with an average of 97.31% and reduces the noise image density up to 12.76% comparative to the existing models.