Deployment strategies for lightweight pavement defect detection using deep learning and inverse perspective mapping

The high cost of pavement detection equipment has constrained its application. Existing pieces of lightweight detection equipment still face the problem of integrated installation, failing to reduce detection costs. Therefore, this paper presents a method for lightweight equipment deployment and installation, where the vanishing point is computed initially using a deep learning approach. Then, based on the vanishing point, the inverse perspective mapping method projects raw pavement surface images into the ones in the “bird's-eye view”. A numerical experiment is used to demonstrate the superiority of the proposed solution. Experiment results indicate that the proposed solution has a 95% accuracy in the vanishing point detection. The influence of deployment conditions on the error of the “bird's-eye view” is