Researches On Segmentation And Recognition Methods Of Pulp Fibre Images

The geometrical feature of pulp fibre such as length,width and crimp, have great influence on the paper qualities. So the measurement of the pulp fibre is significant to product quality control. How to measure the related features of fibre have important meaning in controlling and improving the quality of paper. And that the detection of morphological parameter of fibre first must extract the single fibre. This paper discussed the segmentation and recognition method of pulp fibre in detail.Recently, the researches on the feature extraction method of linear object at home and abroad have come down to many domains. The most used method include of: the method based on edge detection,curve tracing,image transform,template matching and framework extracting. These methods have some applied values, but for the efficiency,precision,universality,anti-jamming and some other aspects also have many work to do.This paper first improved the denoising and contrast enhancing method of fibre image. Use for reference the new image processing method-the Curvelet theory, and combine the feature of pulp fibre image, This method aimed at the direction feature of image, used the orientation performance and high efficiency edge express ability of Curvelet transform sufficiently. Validated that this method can detect the edge of complex image which were low contrast,high noisy,bifurcated or overlapped, by the experiment of edge detection of pulp fibre images.This paper farther discussed the recognition method of generic linear object. Concretely explained the separate and recognize method of interlaced fibres, then combine the width information of fibre to recognize interlaced fibres or bifurcated fibres. The experiment validated that, the image pretreatment method based on Curvelet transform obtained sufficient effect on pulp fibre image segment and recognition.