Research On Gear Image Stitching Algorithm Based On Machine Vision

The demand for machine vision and digital image processing in the field of scientific research is increasing today.In the daily work,only by adjusting the focal length of the camera to obtain a wide viewing angle of the image,as well as more complete image information.The resolution of the camera is fixed,the photos will be affected by the lens parameters,resulting in a certain distortion,the image obtained by this way cannot meet the actual needs of the work.In order to solve the imaging equipment cannot shoot with a resolution and scope of vision problems,this paper based on the theory of digital image processing technology is proposed for the middle and small modulus spur gear digital image mosaic method.Spur gear and small modulus for selection of digital image mosaic image object,the gear obtained were preprocessed using digital image processing technology for the extraction of the edge pixels of gear image,then using the software to calculate the coordinates of the center gear and rotation angle,finally for the image registration,the mosaic of gear digital image.The machine vision technology,image mosaic technology and development trend are introduced briefly.Several typical image filtering methods and image edge detection operators are introduced in detail.According to the theoretical analysis and experiment,the Gauss filter is selected to preprocess the image.The Sobel operator is used to calculate the gradient amplitude of the image,and the edge tracing method is used to extract the contour edge of the image.This paper introduced a novel sub-pixel edge detection method based on sub-pixel edge location method.The three order spline interpolation method is used to segment the image to obtain the sub-pixel edge points,and then the edge is extracted by the Gauss surface fitting to get the sub-pixel boundary.On the same image mosaic method based on the comparative analysis,the mosaic method based on feature extraction,Harris feature gear digital image corner,and uses an improved normalized cross-correlation algorithm for coarse matching of feature points,using RANSAC method for purification of coarse corner matching,accurate matching points are obtained group,the generated affine matrix,obtain two images transform model.Based on image matching edge pixels in the first extract the image edge contour gear based on gradient amplitude using edge tracing method,and then through the least squares method to extract gear rough center and gear tooth extraction region top data points in the theory of constraints on the radius of gear gear precision circle fitting center after registration of tooth profile corresponding to the distance formula,the optimal rotation angle,complete image stitching.The experimental results show that the method based on the edge of the image contour can be used for the high precision stitching of the gear digital images,with very broad application prospect.