Research On Intelligent Vision Measurement System For Large Size Parts Based On Precision Positioning Platform

It is difficult to carry out full-scale visual measurement for the geometry of large-scale planar type workpieces.This topic proposes a high-resolution dual camera based on the high efficiency of visual recognition,the high precision of visual measurement and the high precision of motion control.The low-level coarse and fine two-level visual positioning plus motion control measurement method realizes the intelligent high-precision measurement of large-size parts through visual recognition,positioning and movement of the motion platform.After analyzing the typical contour features of planar parts and the characteristics of dual camera field of view.The intelligent measurement process of large-size parts have been designed.According to the selection of image acquisition and motion control module,the soft and hard structure of the system has been designed.The visual intelligence measurement hardware platform has built,and the overall architecture of the software is designed according to the measurement process and the design of each functional module has been completed.This paper studies the contour extraction method of two-level positioning,analyzes the pixel-level contour method of mathematical image and Canny operator to extract binarized image,and proposes an edge detection method based on Otsu adaptive adjustment of Canny operator high and low threshold..In order to obtain accurate corner positioning results,the image is sub-pixel contoured by Gaussian fitting,and the circular measurement accuracy of the calibration plate under pixel and sub-pixel conditions is compared.Based on the research of intelligent measurement method based on primitive recognition,the contour classification measurement strategy have been designed.In order to realize the rotation,extension and displacement invariance of the part measurement path,an automatic path selection algorithm based on control point centroid vector has been proposed..Data fusion analysis had performed on the results of motion control platform positioning and visual positioning on how to obtain the final result of mobile vision measurement.In the aspect of contour processing algorithm and coarsepositioning,the contour segmentation and classification are studied.A two-step contour segmentation algorithm based on Ramer approximation have been proposed.Based on the size of the dual camera field of view,the segmented contour is based on minimum binding.Rectangular(MBR)classification;in terms of fine positioning,in order to achieve automatic fitting of different types of contours,type identification of segmented contours,and contour fitting based on Tukey weight function,rectangle fitting and Arc fit.The calibration system and calibration experiment had been carried out on the measurement system.The hand-eye calibration based on homography had performed for the coarse positioning module,and the conversion relationship between the global image coordinate system and the motion platform coordinate system have been determined.The pixel-equivalent single-camera calibration method had used to determine the ratio of pixel size to physical size for the local low-level camera.Finally,the precision verification experiment of large and small blocks and the intelligent measurement experiment analysis of large-size parts have been carried out on the system,which proves that the system can automatically identify and measure the parts,and the accuracy of the measurement system is above 0.05 mm.