关键词: 摄影测量, 高清工业相机, Cascade R-CNN, 隧道裂缝识别, 深度学习, 轨道交通

Abstract: Cracks have always been the key detection object of tunnel diseases, but it can only be found by the naked eye and then recorded through the traditional manual inspection. The accuracy and efficiency of manual identification completely depend on personal experience judgment, without the assistance of information technology, so the identification accuracy of operation efficiency needs to be improved. In order to solve the above problems, with the help of high-definition industrial camera imaging by high resolution, fast acquisition speed and other characteristics, the high-definition industrial camera is deployed on the rail car to obtain the tunnel surface crack disease information, which greatly improves the efficiency of tunnel crack recognition, and improves the recognition accuracy by 0.2 mm. At the same time, the optimized cascade R-CNN algorithm is integrated to train tunnel crack samples under supervision. At the same time, we develop a set of crack disease identification route from hardware data acquisition, data processing software and data management platform, which really broke the technical barriers of slow identification, low accuracy, relying on experience and difficult management.

Key words: photogrammetry, high-definition industrial camera, Cascade R-CNN, tunnel crack identification, deep learning, rail transit