Study On The Defect Model And Detection And Recognition Technology Of Multi - Type

After yarn spinning process, the foreign fiber is easy to produce breakage yarn, defects yarn and color difference yarn, which will have serious impact on the yarn quality and result in textile product defects, so many colleges and Research Institute are in the research of foreign fiber detection technology. Currently, cotton foreign fiber detection is mainly detecting single foreign fiber appeared in the visual field, while the detection about many types of foreign fibers appear near the detection point is not so well. However, when the surface quality is not high, there is no need to remove all the different fibers, which only need to remove the larger ones. Based on this research, how to realize the optimization of the multi type fiber co existing conditions are studied, and the main contents are as follows:A new method of dynamic identification of cotton foreign fibers based on mean shift segmentation was proposed in this paper.Using the mean shift method, the original image of the different fiber was segmented, and the two-valued image was obtained;Then, the two-value image was extracted and 12 feature parameters were obtained, and using the hierarchical clustering method and the clustering results by exhaustive method to reduce the dimension of different dimensions of 12 characteristic parameters, and at last we got the 8 effective parameters of the different fibers which were R means, B means, G mean, contour torque, roundness, consistency, entropy and energy.In order to eliminate the fibers more effectively, they established the model of foreign fibers in the yarn faults. Through the tracing investigation of a number of standard sample, they collected the relevant data, and looked the length and area as the independent variable, the defect as the dependent variable, and combined with the linear regression theory to establish the regression equation of different fiber defects. The equation of regression coefficients, the model fitting degree, Durbin-Watson, standard error, Cook distance were significantly tested, and the regression equation would eventually be together to constitute the yarn faults model. When the fiber detection equipment recognized the foreign type of profile fibers, they put the calculated profile fiber size into the corresponding regression equation, and drew the defect points, and compared the size, and then obtained those foreign fiber which are more dangerous and preferentially eliminated.In order to prove whether the model is correct, they had carried on the spinning experiments. After they compare the actual defect from the experiment with the theory of defects from the equation, they can know that the prediction accuracy has reached more than 95%, and the different fiber yarn faults model is correct and effective.The result of the study shows that exists in the multi class specific fiber, yarn model based on optimization selecting method is feasible and for other related fields appear to optimize the culling of multi class object detection has higher theoretical value and practical value.