The 3-D Measurement Of Sheet Forming Work-piece Based On Computer Vision

With the rapid development of modern industry and technology, the 3-D contourmeasurement technology is high required more and more by all kinds of industries. The 3-Dcontour covers all the shape information of the object so that the object can be understoodwholly and deeply through it. The 3-D surface measurement technology is an important andbasal technique for scientific analysis, industry control, bioengineering, biomedicine, materialscience and so on. Especially with the rise of rapid prototyping technology and reverseengineering, the 3-D contour measurement technology is required more highly. And the speedand precision of measurement are more and more important for industry detection, qualitycontrol, computer vision, etc. The structure light used in feature extraction and matching of Non-contact measurementof sheet metal parts is the single direction texture presently. This paper studied the cameracalibration, object segmentation, grid texture feature extraction, stereo matching of 3-Dmeasurement with computer vision with grid texture,which is similar with structure light.During the study, the computer vision, wavelet analysis, method of mathematical morphology,method of image segmentation and genetic algorithm are combined in this paper. Firstly, this paper constructed 3-D measurement system based on computer vision. Thenit makes use of computer vision technology fruits, and calibrates work-piece with the methodof V shape calibration based on two-step method. By V shape calibration block, inside andoutside parameters have been computed then regard them as initial value used in cameracalibration with radial distortion. After the both images are snatched by two cameras they are filtered by wavelet transform.The discrete points of contour are extracted with the method of partial region search and fittedby cubic B-spline curve. As a result, the object is segmented from background and theredundant information is wiped off. And then the discrete points of grid texture are extractedwith wavelet transform, are grouped vertical and horizontal points with mathematicalmorphology, are fitted by cubic B-spline curve and form continuous curves which embody thetrue shape of grid texture. Then an iterative algorithm for parameter correction of B-splinecurve which ensures orientation precision of edge curve. A method of 3-D reconstruction offree form curve is put forward whose elementary is B-spline curve. First, matching parallelcomputing of B-spline curve is realized according to the geometric invariants of vision withthe help of genetic algorithm; second, the free form curve nodes are restored according to III吉林大学硕士研究生学位论文epipolar geometric constraints; last, they are fitted double direction with B-spline curve andthe fit error is corrected through iterative algorithm. Because of the double directiondistribution of grid curve nodes the iterative times is reduced and the reconstructed precisionisimproved greatly. Saddle surface, wave shape surface, sphere shape surface and dish shapesurface have been analyzed in experiments, and explained their error and reason. Several conclusions could be obtained in the measurement of sheet metal parts: 1. The 3-D measurement system based on computer vision is constructed. The methodof V shape camera calibration based on two-step method is put forward on the base of thetraditional camera calibration and compositive consideration all kinds of factors that influencecalibration precision. The stability of camera calibration is increased through differentaltitudinal control points on the V shape calibration block. By calibration, inside and outsideparameters have been computed then regard them as initial value used in camera calibrationwith radial distortion. This method improved the precision of camera calibration. 2. A denoising method of grid texture image is put forward according to the multi-scalecharacteri...