Research Of Pulp Fiber Image Segmentation For The Flow Type Fiber Detector

People’s daily life and work can not do without paper. The quality requirement of paper is also increasingly higher. The quality of paper is decided by all kinds of pulp fibers’ morphology parameters and pure quality. The paper mill’s automation degree is increasingly day by day, and its production is rising rapidly. The manual inspection of fiber quality is time-consuming, and its cost is high. The detection of pulp fiber morphology parameters and features is no longer stay in the level of manual inspection. The research about the method that detecting the morphology parameters of pulp fiber quickly and accurately is practical and significant.The flow type of pulp fiber detector is the equipment detecting geometry parameters of pulp fiber. It is for the pulp through the appropriative flow channel. The equipment should be operated simply and its testing result is accurate. The detection system is composed of a flat channel, CCD camera, ring light source, pipelines, chassis, connectors and other components. This thesis combines the actual requirements of detection for flow type of pulp fiber morphology parameters. It details the mode of image acquisition and maters need attention. Pulp fiber detector is specifically for the analysis of fiber image processing. Before detecting the morphological parameters of pulp fiber, the first thing to do is accurate segmentation of image edge. This thesis is based on the development and study of image segmentation algorithm for the pulp fiber detection, and it masters the mainstream algorithm. Moreover, a variety of fiber images are collected with the flow type of detector, and they are analyzed and compared. Pulp fiber image segmentation method in subpixel level is tested, and it is proved that being suitable for the flow of pulp fiber image segmentation and processing.Firstly, the different methods of image filtering for fiber image are compared. After the mean filtering, the pulp edge in pixel level is extracted from the pulp fiber image by Roberts, Sobel, Prewitt, Canny, Laplacian, Shen, Deriche edge operator. Compared with the results of the edge extraction for clear fiber, fine fiber, fuzzy fiber, cross long fiber, Canny operator is selected. On the basis of this, the chain code location is realized. Each edge pixel adopts quadratic curve fitting. Calculating extreme point after fitting realizes the subpixel edge location. In order to test the effect of location, the results of quadratic curve fitting and Gaussian quadrics fitting are compared. The two effect are similar, but the calculation of curve fitting is relatively less and faster. It is suitable for detecting the edge of pulp fiber images.