| Most domestic sheepskin tannery companies still cut the sheepskin by hand.The labor intensity of workers is high and the working environment is poor.The enterprises are facing increasingly severe recruitment and employment difficulties.In recent years,automation technology and computer vision technology have become increasingly widely used in the domestic leather industry.The automatic edge cleaning of sheepskin can reduce the labor intensity of workers,relieve the pressure of recruitment and employment of enterprises,and improve the production efficiency of sheepskin production and processing enterprises.As an important link in the entire leather production and processing,the study of sheepskin automatic edge removal technology is of great significance for improving the automation level of the leather industry.The contour shapes of sheepskins are different,and the parts that need to be trimmed and cut are also very different.Therefore,effective contour extraction and feature extraction of the cutting parts of sheepskin are the prerequisites for automatic sheepskin trimming.In this paper,based on computer vision,the algorithms for preprocessing,contour extraction and cutting part feature extraction of sheepskin images are studied.The main research contents of this paper are:(1)Research on sheepskin image preprocessing algorithmBy improving the global filtering of the traditional filtering algorithm,a pixel-level selective denoising algorithm was proposed,which modified the gray value according to the distance relationship between the neighborhood median value,the mean value and the gray value.The experimental results show that the improved algorithm has better anti-noise ability and edge protection performance.(2)Carry out research on sheepskin contour extraction algorithmsBy comparing and studying the image segmentation algorithms based on edge and threshold,and combining the characteristics of sheepskin image,a sheepskin contour extraction algorithm is proposed.The Otsu threshold algorithm was improved by optimizing the selection of thresholds in image blocks and accelerating the search of thresholds for image segmentation.Then,the defect edge was repaired and connected by morphological closure operation.Finally,the target contour was extracted by combining the boundary contour tracking algorithm of topology structure analysis and the area threshold method.The experimental results show that the sheepskin contour extraction algorithm can extract more accurate and complete sheepskin contour information,and has good anti-noise performance.Compared with the traditional Otsu algorithm,the segmentation rate of the improved Otsu algorithm is increased by nearly 160%and the false segmentation is reduced by nearly 3%.(3)Research on sheepskin feature extraction algorithmsThe research compares shape description algorithms based on global contours and structural contours.Aiming at the sheepskin features,a sheepskin feature extraction algorithm was proposed,which constructed the structure contour according to geometric relations,used geometric parameter description operator to describe the image element and used threshold model to extract the target pixel.The experimental results show that the algorithm has strong adaptability to sheepskin,it can extract the contour information of sheepskin that needs to be cut from the contour information,and manual extraction of comparative data shows that the algorithm has a correct extraction rate of up to 80%and a high cutting accuracy.(4)Industrialized application experimentAn experimental platform was built using industrial cameras.The experimental results show that the outline and feature extraction algorithms proposed in this paper are universal and feasible for industrial applications. |
