Research On Algorithm Of Subpixel Level Corner And Edge Detection

In the field of computer vision and machine vision,the extraction of image features is very important,and the corner and edge point is a very important part of the image,contains the image of extremely important information,determines the shape of the image of the target,detection and extraction of corner point and edge point information can effectively reduce the amount of calculation,improve the speed of calculation.Therefore,the accurate detection of diagonal and edge information is of great significance in the fields of image matching and segmentation,target recognition and tracking.In this paper,several traditional pixel level corner and edge detection algorithms are studied,and their advantages and disadvantages are analyzed.On this basis,a subpixel level corner and edge detection algorithm is proposed.First of all,in the study of diagonal detection algorithm,four classical corner detection algorithms,namely Moravec corner detection algorithm,SUSAN corner detection algorithm,Harris corner detection algorithm and Shi-Tomasi corner detection algorithm,were compared by experiments to prepare for the sub-pixel accuracy corner detection in the later stage.Finally,Shi-Tomasi corner detection algorithm is selected to initially extract interest points.In the research of edge detection algorithm,five kinds of classical pixel level edge detection operators:Roberts operator,Sobel operator,Prewitt operator,Gaussian Laplacian operator and canny operator were compared,and finally the Canny edge detection algorithm was selected to extract the initial point of interest.Secondly,most corner and edge algorithms adopt the method of differentiation,resulting in poor anti-noise ability.In order to effectively improve the anti-noise ability of the algorithm,the method of integration is considered.For an ideal 2-D L-shaped corner point,consider a model with four parameters:the vertex coordinates of the corner,the direction of the corner,the Angle value,and the brightness values on both sides of the edge.In order to calculate all parameters of the model,six luminance integrals are used.Finally,through the verification of the model,the corner and edge points in the image can be refined effectively.Based on this model and the classical pixel level algorithm,the coordinates of corner points and edge points in the image can be refined rapidly with sub-pixel accuracy,and the false detection rate can be effectively reduced.The use of integral characteristics increases the ability to resist noise.Through the verification of the model,some interfering factors can be eliminated effectively.The effectiveness of the proposed method is verified by experiments.