Research On A Measuring Method And System For The Large-scale Curved Surface Based On Machine Vision

In recent years, the large-scale curved surface has been widely used in theautomobile industry, shipbuilding, aircraft shape etc. Therefore, studying the three-dimensional measurement methods for large-scale curved surface is important forensuring the smooth development and production of major equipment. After acomprehensive review on the indoors and overseas current situation and developingtendency of three-dimensional measurement methods and systems, a three-coordinateaided target measuring method using for the large-scale curved surface based onmonocular vision is presented in this subject.The main work and creative research in this article are as follows:1The principle of monocular vision measurement for large-scale curved surfacewas described. The mathematical model of the measurement system was established,so the coordinate of mark point was calculated. Then realized transfer station measuredin the range of effective viewing area for the camera. By solving the optimizationobjective function with constraints we could get the optimal matrix and achieve3Dregistration. After deducing the coordinate calculation process of measured points, wecould complete the entire measurement for large-scale curved surface.2The process and practical results of the image processing technology inmonocular vision aided target measurement were given. Firstly, the image wassmoothing filtered by Gaussian and enhanced by mathematical morphology, makingiterative threshold and geometry-based feature extraction to achieve the accuratesegmentation of the mark point area; Then using Canny edge extraction and leastsquare ellipse fitting algorithm we accurately striked the coordinate of mark point;Finally, based on the elliptical error evaluation of improved algorithm was proposed,verifying the stability and accuracy of improved image processing algorithm throughexperiment.3The error of monocular vision measurement for large-scale curved surface was analyzed. Firstly, the influence on the accuracy of hardware was analyzed. Then theerror caused by offset and the non-coplanar of mark point taking the deviation fromthe point to target plane as the goal was analyzed. Researched on the formation ofdepth of field, and derived the relationship between the image plane coordinate anddepth of field, and the error correction through the spline sample method was put up.Finally, the radius compensation error for probe was elaborated.4The overall structure for measurement system was designed. Firstly, thedevelopment environment was analyzed. Then we designed the system calibrationprocedure, image processing program, coordinate calculation, error analysis program,and the user operation interface.5Measurement experiment was also carried out according to buildingexperimental platform, then by means of analysing and comparing the experimentalresults with high precision instrument measured the same surface, finally we couldvalidate that the measurement method was feasible and design of the measurementsystem meets the requirements of the theory and measurement accuracy.