Research On On-Line Geometric Measurement Algorithm For Thin Part Based On Machine Vision

With the development of the automatic industry,automatic devices are used not only in the process of the production,but also in the process of the quality detection.The quality of the products can be converted to visualized data,which ensure the fast,efficient and stable detection and statistical analysis.Comparing with the traditional manual operation,automatic detection devices based on machine vision has great advantages in efficiency and long-run cost.But the problems of precision and speed have to be solved.In this thesis,a complete set of algorithm based on machine vision running on the smart camera is proposed,which includes image preprocessing,detection of blocked buttons and subpixel measurement.In the image preprocessing,the ROI region is located quickly by operations include image filtering,threshold segmentation,edge detection and projection integration.The algorithm can detect blocked buttons the blocked area of which is above 10%.Subpixel edge is acquired quickly through two steps.Firstly,edge tracking algorithm based on connected domain analysis is used to locate the edge of the target roughly,and then the sub-pixel edge detection algorithm based on Zernike moments is used to locate the target precisely.The parameters of contours are calculated using the subpixel edge.Finally,the algorithm is ported to the DSP platform and optimized.Experiments on the sample of buttons show that the algorithm gets the geometric parameter at the accuracy of 0.12 mm and recognizes the blocked buttons at the accuracy of 96%.The result also shows the algorithm has good adaptability of light intensity and field orientation adaptability.After optimized,the algorithm runs on the smart camera at the speed of 25ms/fps.