The Study On Object3D Reconstruction Theory Based On Image Sequences And Application In Inspection Of Cable Surface Defects

3D reconstruction technique is the important areas of Computer vision technology andcomputer graphics cross, and widely used in robot navigation, aerospace, medical and otherfields because it can restore three-dimensional the scene information. In order to achieveCable-stayed bridge ropes defect’s location and visualization, the3D reconstruction techniqueis applied to ropes detection.This paper focuses on the crucial technology of three-dimensional reconstruction basedon the sequence of video images, and applies these technologies to the reconstruction ofbridge cable. The principal works and research achievements include following aspects:1.In terms of video preprocessing, with the constant motion of snake-like robot(such asrotation, translation and zoom motion),it will cause the acquired video image sequence blurand judder, not to facilitate the observation and analysis. In the stage of video datapretreatment, this paper applies gray projection of electronic image stabilization to obtain astable video signal having a higher resolution.2.In the camera Calibration, First, introduction and analysis conventional Cameracalibration method，then according to the practical application scenarios，this paper adoptsZhengyou Zhang’s plane calibration method to calculate the initial value of the internalparameters and external parameters of the camera. The intrinsic parameters of the camera areassumed to be same with snake-like robot climb, external parameters are obtained byestimating the fundamental matrix F and the essential matrix E through enough matchingfeature point pairs.3.In term of features feature extraction, firstly，introduced the principle of traditionalHarris and SIFI algorithm, and analyzed their shortcomings, and then proposed a improvedSIFT algorithm for feature extraction based trace of Autocorrelation matrix of the image, thealgorithm reduces feature descriptor dimension numbers keeping with higher precision featureextraction. In term of feature matching, obtain quickly a crude match in the unknown epipolarconstraint using the nearest neighbor Euclidean distance as the metric, and Best Bin First(BBF) as the search strategy, then apply RANSAC algorithm to estimate the fundamentalmatrix F, calculated limit constraint consistency constraints, while taking advantage ofConsistency constraints, uniqueness constraints, Mutual corresponding constraints eliminatefalse match. Finally, experiments show that the matching algorithm to ensure a better matchaccuracy and shorten the matching time. 4.In cable three-dimensional reconstruction of applications, combined Harris and SIFTalgorithm to detect more cable feature points, conducive to the subsequent feature matching.The paper achieved cable reconfigurable experimental, and the experimental ‘s resultsverified the feasibility of rope three-dimensional reconstruction based on binocular visionsystem, and laid the foundation for three-dimensional reconstruction in cable detectionapplications.