Train Mounted Tunnel Surface Image Acquisition And Disease Detection Technology

With the rapid growth of subway operation mileage,how to ensure the safe operation of tunnel infrastructure is particularly important.Tunnel lining is an important part of the tunnel infrastructure.The long-term use of tunnel lining results in a large number of cracks,water leakage and other diseases on its surface.These diseases seriously threaten the safety of subway operation.In order to detect the tunnel surface diseases efficiently and accurately.In order to be able to detect tunnel surface diseases efficiently and accurately,a tunnel surface image acquisition system based on linear array camera is designed.Then on this basis,according to the characteristics of the disease in the collected image,the automatic detection algorithm for the tunnel disease is further studied.The tunnel surface image acquisition system is mainly composed of four parts: image acquisition device,synchronous control device,slave industrial computer and power supply and distribution system.The linear camera and laser light source are integrated in the image acquisition device,which reduces the tedious process of debugging to meet the task requirements of subway tunnel inspection.The collected tunnel surface image is stored in the industrial computer and automatically detected the tunnel diseases.Aiming at the tunnel disease detection algorithm,the research is carried out from two aspects: digital image processing algorithm and deep neural network algorithm.In the deep learning algorithm,firstly,a variety of tunnel target detection data sets containing 9661 images are constructed.Then use the improved Cascade R-CNN target detection algorithm to detect the tunnel target.The algorithm enhances the detection ability for small targets and can more accurately detect a variety of diseases in the tunnel.Since the target detection algorithm can only select the target frame,this article then uses a digital image processing algorithm to segment the crack disease with pixel-level accuracy.In the digital image processing algorithm,the crack enhancement algorithm is first designed to make the crack characteristics clearer.Then use a multi-level fusion filtering algorithm to fuse the pixel information and gradient information of the image to gradually remove scattered noise,large-area block noise and strip noise very similar to cracks.Finally,a seed connected algorithm is used to extract the complete crack area.At this point,the fusion of the two algorithms is completed,and a variety of tunnel targets are detected,and the crack disease is completely segmented with pixel-level accuracy.Several image acquisition experiments were carried out in the laboratory and a subway in South China.The device can acquire the tunnel surface image with the accuracy of 0.2 mm/pix at the speed of 20 km/h.The image acquisition system is stable and safe.The algorithm designed in this paper can recall 96% and 98% of the detection rate of cracks and water leakage,and can segment the cracks from the image with pixel level accuracy.It can be seen that the image acquisition system and the tunnel disease detection algorithm have certain practical engineering application value.