Contour Feature Detection Of Cutting Tools Image

NC tool parameter detection technology is widely used in tool manufacturing and NC machining field, and there is great significance to guarantee the machining accuracy, improve work efficiency and reduce the production costs. In recent years, based on computer vision the tool detection technology has developed rapidly, and in the visual tool detection, image feature detection tool is crucial to improve the detection accuracy, detection efficiency and testing automation degree. This thesis focuses on the tool image contour feature high accuracy real-time extraction algorithm for research.The detection methods of the tool image contour feature are studied in this thesis. The key technology of sub-pixel edge detection, contour extraction, corner detection and tool geometry parameters measurement are thoroughly analyzed, and realize the measurement of the tool basic geometric parameter.A sub-pixel edge detection algorithm with Gauss interpolation is optimized in this thesis. According to the interpolation method of mathematical theory and sub-pixel interpolation method theory, optimize sub-pixel edge detection algorithm with Gauss interpolation, realize the image edge tool of the sub-pixel location, and can effectively improve the accuracy of image edge detection. Through the comparison of experiment result between the new algorithm and the traditional sub-pixel edge detection algorithms, it is demonstrated that this algorithm has higher sub-pixel edge positioning accuracy.A contour extraction method based on identification of watershed Snake model is improved in this thesis. Extracting the image contour that the target shape is complicated, according to the classic Snake algorithm, it is easy to appear the concave contours which is difficult to approach, simultaneously need artificial to draw the outline of the initial contour, which is a time-consuming project.combined with the advantages of watershed algorithm proposed setting method of Snake contour model identification algorithm based on watershed, by controlling the initial contour model, avoiding excessive segmentation, reducing the number of iteration to get faster contour extraction rate and good adaptability. The experimental results show that this algorithm can quickly and accurately extract the outline of the model.A detection method for fast high precision corner is improved in this thesis. The tool image contour through extraction also exists some residual angular corner, in order to obtain complete tool feature information, through comprehensive consideration of Harris corner detection algorithm, Susan corner detection algorithm and the characteristics of tool image, this thesis proposes a fast and accurate detection method which has good robust to noise, realize the corner detection, thus complete the integrity test of tool feature. The other corner detection algorithm with the improved corner algorithm is compared:show that this method improves the corner detection accuracy, more in line with the requirements of high precision image measurement.In this thesis, according to the sub-pixel edge detection algorithm, contour extraction algorithm and comer detection algorithm, studies the measurement of geometrical parameter measurement method based on least squares fitting and realizes the tool basic geometric parameter measurement. The experimental results show that this method has high measuring accuracy and the operation time is shorter.