Research On Tunnel Image Acquisition System And Crack Intelligent Identification Method

In recent years,with the rapid development of urban rail transit in China,the total length of tunnels ranks first in the world.Due to the influence of various factors such as construction,temperature,load,etc.,diseases on the surface of subway tunnels may occur.Among them,the crack on the surface of tunnel is one of the most common diseases,which poses a threat to the safe operation of subway.Therefore,crack detection is an important task of periodic inspection of subway.At present,the detection of subway lining cracks is still based on manual inspection.This paper develops a tunnel lining image synchronous acquisition system based on multi camera,which can quickly collect high-quality tunnel lining images in the skylight time.At the same time,a set of intelligent crack recognition method based on deep learning and traditional image processing is proposed to realize the intelligent detection and risk assessment of crack diseaseThe image acquisition system of shield tunnel lining designed in this paper includes plane array camera and light source integration module,camera bracket,large capacity industrial computer and power supply system,and the following improvements are made:firstly,twelve groups of camera modules are designed to realize full section acquisition;secondly,the method of serial communication is used to realize the synchronous acquisition function of all camera modules;finally,the mutual conversion between the image pixel size and the physical size is completed.The maximum speed of the system and the frame rate of the camera are calculated.The whole tunnel image multi vision acquisition system can be flexibly installed on the track mobile inspection platform to realize the automatic acquisition of tunnel crack image.Intelligent crack recognition methods mainly include: crack recognition algorithm based on edge detection network and image post-processing and risk assessment scheme of crack area.This paper constructs a subway tunnel crack image sample library,makes pixel-level crack edge detection data sets.Based on the classical edge detection model holistically nested edge detection(HED)and bi-directional cascade network(BDCN),the crack edge detection network is optimized and improved.Based on the network structure of HED,the feature pyramid and hole convolution are introduced to realize the automatic recognition of crack image.A post-processing algorithm based on digital image processing is designed to further extract the cracks.An adaptive threshold segmentation algorithm is used to obtain the binary image of the cracks.Then the connected domain filter is used to eliminate the speckle noise.Finally,the morphological closed operation is used to fill the holes in the cracks and smooth the edge of the cracks.At the same time,based on the fracture skeleton extraction model,the intelligent classification and recognition of linear fractures(longitudinal fractures,oblique fractures,circumferential fractures)and complex fractures are realized.According to different fracture characteristics,a typical parameter measurement algorithm is designed,and a fracture area risk assessment method is proposed.The multi vision image acquisition system developed in this paper has carried out synchronous acquisition and algorithm verification experiments in the section between the simulated subway tunnel and the actual subway tunnel.The results show that the system can acquire crack image with accuracy of 0.2mm,and the whole system runs stably.The designed crack detection algorithm can achieve pixel level detection,the IOU can reach 0.3912,the accuracy rate is 83.79%,and the recall rate is 78.77%.It can be seen that the image acquisition system and crack detection algorithm can provide better technical support for the actual tunnel crack detection.