Analysis And Identification Of Cotton/Hemp Fiber Cross And Longitudinal Morphology By Digital Image Method

Since 1990 s,domestic and foreign scholars have done a lot of work on the measurement and recognition of fiber characteristic parameters by using image processing method.As a basic testing method,manual testing takes a long time and costs a lot.The testing results will be affected by subjective emotions,fatigue and other external factors,resulting in testing errors.The image method can measure the characteristic parameters of the fiber and analyze and recognize them,which has the advantages of fast operation speed,high efficiency and accuracy.Based on the research progress of researchers at home and abroad,this paper proposes a new algorithm,in which the cross-section image of cotton / hemp fiber is preprocessed by the edge enhanced image fusion algorithm,so as to weaken the image background,enhance the contrast between the target and the background.The edge enhancement image fusion algorithm first grays the original image,then extracts the fiber edge by using edge detection operator,then fuses the gray image with the edge image by using image fusion algorithm,and finally extracts the parameters of the image.This method reduces the influence of resin,knife marks and other impurities on image quality,enhances the edge of the fiber,and enhances the contrast between the target and the background.After preprocessing,the fibers in the cross-section image will be damaged and conglutinated.To solve this problem,this paper proposes a fiber repair algorithm based on boundary tracking,which uses the boundary tracking function to identify and mark the target fibers,obtains the coordinates of the fiber cross-section edge points,and then finds the start and end points of the repair according to the changes of the coordinates of the edge points,and finally uses the Arc repair fiber.After the restoration,the transverse characteristic parameters are measured.The cross-sectional characteristic parameters of cotton / hemp fiber to be measured include cross-sectional area,perimeter,roundness,maximum diameter and minimum diameter.Because there are many transverse characteristic parameters to be measured and manual measurement is time-consuming,this paper usesthe coordinates of the edge points of the fiber cross-section recorded by the boundary tracking algorithm to quickly measure the morphological characteristic parameters of multiple fiber cross-sections.For the longitudinal image of the fiber,the background separation algorithm is used to preprocess.The background separation algorithm realizes the separation of the target fiber and the background by changing the filter template of the mean filter.The longitudinal characteristic parameters of cotton / hemp fiber to be measured include fiber fineness,medium cavity ratio and twist degree.In this paper,Hough transform is used to detect the longitudinal edge of fiber,and the edge function of both sides is obtained,then the fiber fineness is calculated according to the function.Because the line detection method of Hough transform can't detect the edge of the middle cavity,this paper uses the manual point method to measure the width of the fiber middle cavity,and then calculates the proportion of the middle cavity.When calculating the twist degree,this paper adjusts the sensitivity threshold of edge detection operator and removes small pixels,so that edge detection can only detect one side edge line,then uses the boundary tracking method to record the coordinates of edge points,and takes the coordinates of the first point and the last point to calculate the twist degree of fiber.In this paper,when measuring the transverse and longitudinal parameters of cotton /hemp fiber,the method of batch processing is used.The measured longitudinal and transverse parameters do not belong to the same fiber,that is,the transverse and longitudinal parameters of cotton / hemp fiber cannot correspond one by one.This paper classifies and identifies the two directions of cotton / hemp fiber cross section and cotton /hemp fiber dimension respectively.In the aspect of classifier,this paper chooses LSSVM and BP neural network for training test and analysis.