The Study On Line Detection Methods Of Color Images

The line detection technique is mostly designed based on the edge image.Edge and line features are the most basic features of images,containing the most of the important information of the scene in images.Edge detection and line detection play a crucial role in image processing,widely used in remote sensing image analysis,target recognition and robot navigation.Due to the limitation of scientific technology and hardware in the early stage,edge detection and line detection are mostly designed based on gray-scale images.Color is an important descriptive feature in image processing.Compared with gray-scale images,color images can reflect the color information of the scene,and can describe the details in the scene in more detail and accurately.With the improvement of the capacity of hardware storage data and computing performance,as well as color images being commonly used,the research of color image processing technology has become an urgent task to be advanced in the field of computer vision.In this paper,color edge detection and line detection of color images are mainly studied.Firstly,on the basis of the understanding about the fusion of color information in edge detection,a color edge detection algorithm based on the convex combinatorial form is proposed,achieving the fusion of color information by constructing objective function based on convex combinatorial form.The fusion process of color information is regarded as the linear combination of color vector components with a certain proportion or weight.Then by means of calculating iteratively the anisotropic gaussian directional derivative(ANDD)matrix with unknown weights by singular value decomposition,the maximum singular value of the matrix in the last iteration is the maximum fusion edge value based on the ANDD matrix.Following the maximum fusion edge strength and direction based on the gradient matrix and ANDD matrix are calculated respectively.Through the intuitive comparison of simulation experiments and the qualitative and quantitative evaluation of ROC curve,it is verified that the algorithm proposed in this paper can detect more edge details,as well as having good detection performance and being robust to noise.Secondly,combining the color edge detection algorithm based on gradient matrix and ANDD matrix with the line detection algorithm of gray-scale images,a color image line detection algorithm based on the color edge detector is proposed.First of all,carry out the color edge detection,and next perform clustering operation on the edge image.Then calculate the parameters of the elliptic Gaussian kernel function according to the uncertainty of the best-fit straight line of the clustering combination.whereafter vote using kernel function and get the parameters of lines by peak detection.The algorithm in this paper considers the color information of images and can effectively detect the lost line structure due to the gray-scale processing.Finally,according to the thought about different merging ways of color information in color edge detection,a kind of color line detection algorithms based on the funnel transform are proposed.By applying different merging methods of color information,color line detection algorithms based on the jacobian matrix analysis,the convex combination or ₂L norm are proposed,which mainly include color Canny edge detection,funnel transform,the fusion process of color information and post-processing.A large number of simulation experiments conducted on the York Urban DB data set show that,compared with the straight line detection algorithm based on funnel transformation in gray-scale images,the straight line detection algorithm based on funnel transformation in color images utilizes the color information of image,and can detect more lines describing details in the scene,having better detection performance.