Research On Crack Identification Technology Of Block Masonry Wall Based On Full Convolution Neural Network

In the rapid development of infrastructure construction and buildings in China,more and more concrete and masonry structures are used.At the same time,cracks exist objectively.Crack identification and detection has been a hot topic for scholars.With the development of deep learning and computer vision,there are more and more researches on the application of neural network in the field of civil engineering crack recognition,including roads and bridges.There are relatively few researches on the crack recognition of concrete and masonry walls.At the same time,it is of great significance to the detection of wall cracks in both college research and daily life.Therefore,this topic In this paper,the crack recognition of masonry wall is studied.In recent years,the well-developed full convolution neural network technology is used to carry out the semantic segmentation task of cracks and complete the identification,detection and classification of cracks.This paper starts with the application of the full convolution neural network recognition technology,introduces the theoretical knowledge of machine learning,and then introduces the advantages and applications of the full convolution neural network in semantic segmentation.Secondly,taking the crack image of the block masonry wall collected in the laboratory as the sample database,and making a simple classification for the cracks of the block masonry wall.Then,on the basis of the original image database,artificial pixel calibration is carried out to obtain the same number of calibration image database.In order to improve the accuracy of neural network training results,some image processing is carried out,including mark note elimination,traditional image processing,image clipping,data enhancement sample diversification and other operations.Then it introduces that the neural network architecture used in this paper is based on vgg-19.Then,the whole connection layer and other networks are replaced by the anti roll layer.The FCN neural network model is composed of roll layer,pooling layer,activation layer and anti roll layer,and the ratio of activation function,loss function,weight optimization algorithm,initial selection of weight parameters and learning rate is compared The comparison and analysis of the neural network model,as well as the results of the operation of a comprehensive discussion.Finally,based on the crack segmentation image predicted by the network model,the traditional image morphology processing is carried out,including expansion,corrosion,skeleton operation,and then the crack length and width are calculated based on the pixels of the crack image.Finally,according to the neural network,a simple analysis and discussion are made on the comparison between the fracture classification results and the fracture image.Through the research of this paper,the recognition,detection and classification of the crack image of masonry wall based on full convolution neural network is realized.