Adaptive Feature Extraction Of Cutting Tools Based On FCM

In machine vision and image analysis, image segmentation is very important. Image segmentation is one of the main research subjects in the fields of image processing. Segmentation directly determines the effects of follow-up digital image processing, digital image analysis, and digital image understanding. Image segmentation has high research value. A quick feature testing method of tool image segmentation is very important. The method can help to improve the efficiency of the digital image cutting tools measurement machine and recognition.The method also has a great significance for the image matching, image recognition, and image based modeling.The main contents of this thesis are fuzzy c means clustering algorithm (FCM), contour tracing, and feature extraction. The traditional FCM clustering algorithm has some problems, such as a massive calculation and a slow operation speed, especially a large amount of the data. A fast multi-thresholds FCM algorithm based on histogram correlation constraints is proposed to control the image distortion due to resample. Because of the amount of data in the operation has been reduced the segmentation speed turns faster. Using the fuzzy c means clustering algorithm image segmentation for cutting tools; continue to study the contour tracking, by experiment advantages and disadvantages of the field contour tracking algorithm, and the introduction of the coding technology. Through this profile, calculate the curvature, Harris corner detection, and cubic B-spline curve fitting.These methods have been getting the feature points of cutting tools image.The fast fuzzy c means clustering algorithm cost much less time than the traditional FCM when segmenting gray level image and take less average steps of iterate computing. The algorithm gets acceptable efficiency while ensured the segment effect to make it possible to apply in analysis high resolution images. Segmented images have been followed to got an availability aggregate of points and reduce the rate of producing edge defects when measuring by parallel ways in order to raise the precision of detection. Meanwhile, hybrid programming has been researched by the article.