Research On Concrete Surface Defect Detection Method Based On Fusion Of 3D Imaging And Photogrammetry

The detection of surface defects of large-scale concrete buildings,such as dams,is of great significance to ensure the structural safety of buildings.It is an urgent problem to explore automatic detection methods.In this paper,taking surface cracks as the detection object,three key technologies of concrete building such as three-dimensional modeling,medium-long-range imaging and automatic crack detection are systematically studied.The main work and innovation are as follows:(1)2D-ELAS(2D Efficient Large-Scale Stereo Matching)algorithm is proposed to realize 3D modeling of concrete buildings with photogrammetry fusion.Based on the basic matrix,epipolar deviation and flatness constraints,a stereo matching and constrained map iterative updating method based on FEC(Fast Edge and Corner)feature points,is designed to obtain a reliable set of matching support points.From the support triangle and matching constraints,the two-dimensional Bayesian probability estimation is used to obtain dense point cloud,and the 3D reconstruction of weak texture surface is realized.The high-resolution image mosaic and registration are used to realize the mapping from the crack detection results to the 3D model.(2)The defocus estimation model of circular edge response curve is established,and the defocusing radius estimation algorithm of blurred image is designed.The influence of various noises and motion blur is evaluated.Based on the range focusing model,an optimized calibration method of focusing parameters is proposed,and the automatic focusing imaging system is verified by experiments.(3)The advantages and disadvantages of pixel based and region based crack detection methods are analyzed.According to the theory of Finite State Machine(FSM),a new FSM double edge extraction operator is created,and a crack recognition algorithm based on random forest is proposed.The adaptive threshold method of FSM operator is studied,and the time complexity is analyzed and verified.The detection sensitivity and processing speed of the algorithm are compared and evaluated.(4)A set of 3D modeling and photogrammetry system is developed,including high-resolution imaging,3D modeling,scanning control,image management and fracture analysis.The developed system is used to test the crack detection of a water gate,and the effectiveness of the overall detection system is verified.