Research On Bridge Disease Identification Based On Deep Learning

Bridge plays an important role in the process of transportation,so it is an important work to ensure the safety of bridge structure.There are countless bridges of all sizes in the country.If we rely on manual on-the-spot inspection,the efficiency will be low,the cost will be high and the risk will be high.How to improve the efficiency and accuracy of bridge detection,without repetition and omission,has become a research problem of many scholars.In this paper,a series of research has been carried out for the detection of bridge diseases,the main works are as follows:(1)In view of the low efficiency of bridge disease classification,a bridge disease classification method based on convolution neural network(bd-cnn)is proposed.By building the bridge disease database,selecting and improving the network framework,building the hardware environment,optimizing the training parameters for many times,and finally generating an ideal classification model,the precise classification of cracks,exposed steel bars and pockmarks is realized.(2)In order to solve the problem of background interference in the process of crack feature extraction,which is easy to produce false detection and omission,a crack detection model based on FCN combined with region growing algorithm is proposed.Firstly,the improved alexnet framework is used to complete the rough segmentation of crack region.Secondly,the region growing algorithm is used to further optimize the segmentation results and extract crack features Finally,the detection results are output according to the proposed crack width measurement method.(3)In view of the problem that there are many interferences in the area of missing reinforcement and pockmarked surface under the complex background and it is not easy to extract,the disease extraction algorithms based on projection method and morphology theory are proposed respectively.Through different filtering algorithms and segmentation methods,the background is effectively removed and the morphological characteristics of the disease are retained.Finally,the disease characteristics after noise removal are quantified to facilitate the subsequent safety of the bridge Assessment.