Research On Image Mosaic Of Composite Materials Based On Texture Features

Composite moulded products are widely used in military,civil and other fields.The defects such as crack,pitting and inclusion are easy to occur after the molding of molded products,and there are many texture interferences in the molded products of composite materials,which leads to the problems such as missed inspection,false inspection and inaccurate size detection in the process of manual detection.The detection technology based on machine vision can effectively replace manual detection of defects in molded products.But,as the research object in this paper such artifacts is a conical barrel more pieces,and diameter is large,up and down due to filming moulded products inside a week to get the lining of the overall image,and pictures only a limited range of area,how image stitching would need to be taken for many times and get the complete wall panoramic images.To solve this problem,this paper proposed an image Mosaic method of composite molded products based on gray co-existence matrix texture feature block and SURF algorithm.The main research contents of this paper include:On the basis of in-depth analysis of composite image texture feature,through a variety of texture feature description method of comparison,the method based on gray level co-occurrence matrix of composite material texture characteristics of overall description,analysis of the image noise influence on gray level co-occurrence matrix to extract data,and the influence of the image noise through the median filtering reduced,for the subsequent composite moulded products image fusion splicing and provide high quality images.In order to realize the splicing of composite molded products images,a registration method combining thick and thin matching is adopted.The composite molded products images are processed into blocks by means of equalization method.SURF algorithm was used to extract the image features in the acquired matching blocks to achieve precise matching.Finally,the global homography matrix obtained by the size matching is used to realize the image registration of composite molded products.Experimental results show that the registration method proposed in this paper can realize the image Mosaic of composite molded products with high efficiency and good quality.Considering the stitching images caused by shooting problems appear spell joint dislocation,ghosting,texture,etc.,resulting in a decline in the effect and quality of the image matching,based on the particularity of the composite material moulded products image texture feature,this paper adopts weighted poisson fusion algorithm,after mosaics of the composite material moulded products image fusion processing,make stitching image overlap part can smooth transition,to realize the seamless integration of texture image stitching.Fusion experiments show that the fusion algorithm adopted in this paper does not find any misalignment or patchwork joint in the overlapping region,and achieves smooth transition of pixels between texture images with good fusion effect.