DEVELOPMENT OF AUTOMATED CRACK DETECTION METHOD BASED ON CONVOLUTIONAL NEURAL NETWORKS

摘要

So far, the practically viable techniques for pavement crack detection comprise manual labor. Hence, a significant amount of work and time is needed. In this paper, we established a classifier to detect pavement cracks using Convolutional Neural Network, which has been trained from the pavement images acquired by a line sensor mounted onto a road inspection vehicle. Our network was trained to detect cracks in the pavement images into the ones that contain cracks or no-crack. By analysing the results obtained from our proposed model, we have validated that our approach was successful in recognizing the cracks with high accuracy.