Research On Detection Of Tunnel Lining Crack Based On Faster R-CNN

With the rapid development of transportation in China,the construction of tunnels is one of the most important parts,and the maintenance of tunnels is very important.The detection of tunnel lining cracks is related to the safety of traffic and transportation.At present,the detection of tunnel lining cracks in China adopts the method of manual inspection.With the development of artificial intelligence machine vision,the combination of neural network and deep learning has been widely used in the field of image recognition.In this paper,the relevant theories of deep learning and the current popular target detection algorithm are deeply studied.Based on the R-CNN(Region-CNN),Fast R-CNN and Faster RCNN frames of convolutional neural network,the recognition speed and accuracy of tunnel lining cracks are greatly improved.Based on the classical Faster R-CNN framework algorithm,this paper studies the detection and recognition of cracks in tunnel lining images.Based on the similarity of local and global features of cracks,the tunnel lining crack detection system proposes a crack identification method,which effectively improves the accuracy and leakage rate of identifying micro-cracks in large-scale images.The advantage of residual network is used to reduce the training parameters of the network,and in view of the imbalance of positive and negative samples,the degree of difficulty in distinguishing samples and the distinguishing degree of samples' features,the joint training of focus loss function and center loss function is proposed.Through the analysis of experimental results,the improved Faster R-CNN algorithm proposed in this paper is better than the original tunnel lining crack detection system in detecting the leakage rate and error rate in the lining crack system,realizing the end-to-end effective training.