Research On Measurement System Of Small Workpiece Size Based On Machine Vision

With the development of automation in machining industry, higher requirements are also put forward for dimensional measurement of machined workpiece, since the traditional measurement techniques cannot meet the needs of the current development. This paper studies the small workpiece measurement system based on machine vision, which mainly solves the following problems: low speed of manual measuring, fatigue for long time measuring, low accuracy of measurement, difficulty for automatic measurement and so on.According to the demand of measurement, some hardware devices should be carefully selected first, such as light sources, cameras, lenses and so on, which form a hardware platform of measurement system for calibration and image acquisition. To highlight the boundaries of the workpiece and avoid the reflection of metal surface, the LED annular shadowless light is used to get boundaries of the workpiece. In addition, the models of camera, lens and other hardware devices have been determined, whole hardware platform of measurement system is constructed, and the camera calibration and the image shooting are completed.Matlab8.0 is used for programs, and the whole process is divided into five parts.The first part is image filtering, after comparing and analyzing, median filtering algorithm is used to filter out the noise in the image without destroying the edge information of the workpiece.The second part is the processing for image binarization, and we use histogram threshold method to determine the threshold to distinguish the edges of the image from the background clearly. The third part is edge extraction, and the Canny algorithm is used to effectively extract the edge information of the image, including the information for the edge of workpiece and noise part. The forth part is to eliminate the edge of noise part, and the aggregation algorithm of highlight regional for binary image is proposed in this paper, which first divides all edge pixels into different clusters, including edge pixels of noise and edge pixels of workpiece, then removes the edge pixels of noise by using the difference between the number of the edge pixels of noise and workpiece, and get the edge pixels of workpiece as a result. The fifth part is the segmentation of edge, the mapping segmentation algorithm is proposed in this paper, which finds the inflection points on the edge by using the characteristic of large slope difference between the points before and after the image inflection point, and divides the continuous workpiece edges into sections accurately and effectively. Finally, for the kinds of edge of the workpiece are straight line and arc, each edge is fitting as straight line or arc, and then the length of each edge is got to complete the measurement.Experimental results show that,small workpiece measuring system based on machine vision in this paper could implement real-time, automation and high precision measurement.