Intelligent Identification Of Bridge Structural Cracks Based On Unmanned Aerial Vehicle And Deep Learning

The crack of bridge structure is one of the main preventive diseases in bridge daily maintenance.Surface crack detections of bridge structures provide important condition information and decision basis for condition identification,disease regulation,and safety assessment of bridge structures.To address the problems of the traditional manual detection method,such as high risk,traffic impact and high cost,this paper proposes an intelligent crack identification method for bridge structures based on unmanned aerial vehicle(UAV)and deep learning,an intelligent bridge crack detection method based on UAV and deep learning is proposed in this paper.The intelligent bridge disease detection is realized by using UAV system,deep learning,digital image processing and other machine vision technology and artificial intelligence technology.By using UAV system,deep learning,digital image processing and other machine vision technology and artificial intelligence technology,intelligent bridge disease detection is realized.The bridge appearance detection method proposed in this paper has the advantages of non-contact,low risk and automation,which can improve the efficiency of bridge routine inspection,reduce the risk and cost of bridge detection,and has important scientific research and engineering application value.The main research contents are as follows.(1)Research on the feasibility of UAV bridge detection.This paper studies the specific flight strategy and aerial photography mode of UAV in bridge detection,fully considers the UAV model selection,camera lens selection,shooting angle and other factors,obtains the best shooting distance of different width cracks,so as to determine the best aerial photography distance of UAV system.With the help of RTK module,based on the real-time pose information and obstacle avoidance distance information of UAV,a three-dimensional relative coordinate system is constructed to determine the specific coordinates of UAV aerial image and realize aerial image positioning.An image quality discrimination method based on entropy sharpness is introduced to screen out high quality aerial images.(3)Research on Intelligent crack detection method based on deep learning algorithm and intelligent recognition technology of deep learning object detection algorithm.Using the SDNET crack dataset and other image resources,we established a training image library for the deep-learning,covering 1133 images precisely marked crack aeras.The deep-learning recognition model was trained and established using the Region-based Convolutional Neural Network(Mask R-CNN)algorithm.Based on Mask R-CNN identification model,cracks were automatically identified and located by means of scanning high-resolution concrete surface images in rectangular sliding windows.The trained target detection network model is used to detect the cracks on the tower surface of Hongshan Bridge in Changsha City with a height of 136.8 M.the accuracy and recall rate are above 90%,and the intelligent crack recognition of highrise bridge is successfully realized.(2)Research on fracture shape extraction based on digital image processing technology.An image post-processing method including image binarization,connected domain de-noising,edge detection,crack skeletonization,crack width calculation and other processes is constructed to realize the automatic acquisition of crack shape and width information.Through the accuracy verification test,it is confirmed that the crack width identified by the method proposed in this paper is consistent with the results of the crack measuring instrument,and the absolute error is less than 0.097 mm,and the relative error is less than 9.8%.