Design And Implementation Of A Large Leather Image Stitching And Contour Extraction System

Image stitching technology is a technology that seamlessly splices two or more images that overlap each other.Leather images are generally used in leather processing and other clothing industry.As the animal leather format is large,generally need to shoot multiple times to obtain images,and then stitch the leather images.Large-format leather image contour extraction technology is the use of image processing technology to obtain the contours of the available parts of the leather image.The technology for the leather industry's nesting and ut ilization rate has practical application significance.In this thesis,we mainly focus on the research about the design and implementation of large-scale leather image stitching and contour extraction system.We improve the feature point matching and contour extraction of image mosaic,and further optimize the edge processing method of fur image,and finally realize the large-scale leather image stitching and contour extraction system.The main content of the work is as follows:1.In this thesis,pretreatment and stitching of large format leather image are realized.The image is processed by gray scale,filtering,binarization and morphological processing.In the process of large-format image stitching,this thesis presents an algorithm for splicing the image according to the position information of the checkerboard feature points of the repeated part of the leather image.Through this algorithm,the feature point detection of checkerboard repeat region is realized.According to the matching relationship of the repeat regions,the transformation matrix can be calculated and the image stitching is realized.Compared with the method of Harris corner detection,all the feature points of the two images are matched.The method of this thesis can improve the efficiency of image stitching.2.In this thesis,an automatic threshold calculation method for image is designed.The image adaptive binarization is realized.The method can separate the background of the image from the target leather.A vector angle threshold method is proposed to preserve the contour points in the range of the angle threshold.and remove the points of the smooth area.The method can preserve the image feature,use as few corner points as possible,and finally achieve the effect of using a few numbers of points to represent the contour of the image,and realize the effect of contour optimization.O n the basis of the optimization of the contour point,according to the known actual size of the checkerboard,the bilinear interpolation is used to vectorize the contours of the leather so that the actual dimensions of the vectorized contour are in accord with the actual boundary of the leather.3.The fur area in the fur image is processed.In order to prevent the image of the fur area from making errors,this thesis presents an improved method of binarization threshold process.We Segment the image by setting reasonable threshold,so the fur area can be removed.Experiments show that the introduction of binarization can improve the accuracy of contour recognition.4.In this thesis,we design and implement a large leather image stitching and contour extraction system which can stitch leather images and remove fur area of the fur image while able to identify the leather available area then extract the contour of the area.