Research On Fiber Image Classification Algorithm Of Pulp Fiber Measuring Instrument

With the rapid development of China's industrial modernization and the rapid growth of market competition,the demand for the traditional manufacturing industry is getting higher and higher.The paper is one of the four great inventions of ancient China,once the culture communication history made outstanding contributions,even in the electronic books popular today,the traditional paper and paper products still plays an irreplaceable role.However,in the paper production process,the production of raw materials,production equipment and production process and other issues,in the paper production process there will be some paper disease,it is because of these paper defects exist,have caused great economic losses to the papermaking enterprises.The papermaking enterprises the most energy on paper detection,cause of detection and classification of paper surface defects has been very mature,but the production of raw materials,pulp-Detection and classification is still a blank.The quality of the raw material directly determines the quality of the product,and the detection and classification of pulp fiber is the most direct guarantee for the quality of the paper.Therefore,it is very important to study the detection and classification of pulp fiber.Firstly,background subtraction image and treatment of pulp fiber image,remove the dust on the lens attached to the imaging attachment;then through the comparison of different filtering algorithm,finally selected a template for the 3*3 Gauss filter to filter the image and pulp fiber image;secondly the filtered images are the nine types of segmentation;characteristics of two kinds of morphology and grayscale extracted in the feature extraction;in SVM classification algorithm,the SVM training model,the Gauss kernel function,using a classification strategy one on one,although the better classification effect of SVM fiber,but the classification of the pulp the effect is poor,not up to the result that we want.In the convolutional neural network,in the training of a convolutional neural network,we choose the convolutional neural network structure of three layers,in which the number of neurons in the input layer is set to 9,the output layer neurons number is set to 6.The number of neurons in the hidden layer is analyzed and compared with several experiments,and finally set to 10.Although the classification effect of convolutional neural network is not as good as SVM,the classification of pulp has reached a high level.We first use convolutional neural network to distinguish the different kinds of pulp,and then use SVM to classify thesame kind of fiber and non fiber impurities.Finally,through the study of different kinds of pulp papermaking performance,discusses the influence of different parameters on the morphology of pulp papermaking performance,the papermaking performance of fiber and fiber image classification of the mixture of model prediction verified the relationship between the shape parameters of fiber and the papermaking performance.