Road and Bridge Expansion Joint Crack Detection and Disease Classification Based on Deep Learning and Morphology

Abstract

Roads and bridges play a pivotal role in China's land transportation. However, with the increase in operation time, the concrete in the anchorage area of road and bridge expansion joints will be subjected to fatigue loading for a long time cracking will occur, and the expansion joints will be bulging and other diseases. In turn, it may lead to a major accident. To quickly detect whether there are cracks on the surface around the road and bridge expansion joints and the type of cracks, a road and bridge expansion joint crack classification model was established. The experimental results show that the accuracy of the model reaches 98.97% after three iteration cycles, and the loss value is less than 0.06. Then, morphological filtering algorithms based on the computer vision framework OpenCV were used to calculate the maximum width, average width, and average angle of the cracks. This can be used to categorize the degree of crack damage in road and bridge expansion joints. Finally, through experimental testing, the relative error between the crack width value and the actual value measured by the reading microscope is within 3%.