| Highway tunnel is an important part of public transport facilities in China,but its operation may cause facilities diseases for a variety of reasons,including cracks,seepage and spalling,etc.Tunnel crack is the most widely disease in tunnel diseases.In the early years,people generally use artificial screening tunnel cracks,costly human resources,along with the rapid development of computer technology,many based on computer and image processing technology,but hard to improve on for crack identification accuracy,the effect has been difficult to meet the requirements of industrial detection.Therefore,it is of great significance to find effective ways to identify tunnel crack identification.Deep learning techniques have flourished in recent years.Especially in the deep convolutional network,it has achieved remarkable success in the research area of image classification and recognition.In this paper,based on the depth of the convolution network(CNN)crack image classification recognition technology research of the tunnel,the classical VGG network model is improved and designed my own network,applied to the tunnel crack image classification and recognition applications.It is proved that the limitation of traditional algorithm is to introduce the feature method of automatic extraction of deep learning.Through offline gathering image,and build the tunnel based on the crack image of large quantities of data sets,designed a data set and the means of dividing the network training methods to train the network,and by using various trainings for cross validation.In the end,we compare with many traditional recognition algorithms and demonstrate the effectiveness of the proposed method.In summary,this paper proposes a tunnel crack image classification recognition method based on CNN,compared with the traditional method to obtain the significantly better performance,and get the average recognition rate is as high as 98.7% in the image classification and recognition of tunnel crack. |
PDF Full Text Request