Crack Detector Automation for Roads by Deep Learning

Abstract

A thorough analysis is required for assessment of structural health of road infrastructures. A Computer Vision Technology based analysis of images will help us for better management of roads/ pavements in long run. A Convolutional Neural Networks (CNN) based Deep Learning algorithm can be constructed, which can take in an input image used for pavement classification. In this paper, a python based algorithm is proposed and it is simulated for experimental results using automated system software which detects the cracks on the road. The approach used in this paper is mainly based on full surface reconstruction in 3D pavement. The automated method achieves same accuracy with a reduced cost when compared to manual computation.