Image Edge Analysis and Application Based on Multimodal Analysis

Sub-pixel technology can improve the edge accuracy of the image. In this paper, the sub-pixel technology is studied to extract the edge contour of the image. In this study, firstly, three calibration plate images are analyzed, the outline of the edge of the calibration plate is extracted, and the points are fitted by Matlab software. It is found that the rotation angle of PiC2 is only 0.0775 °, and the correlation between long and short axes is very small. Therefore, Pic2_1 is considered to be a standard image. The image rotation angles of Pic2 and Pice2 are 3.7085 °and-3.3435 °, respectively. According to the scale of 116.86 pixels is 1 mm and the rotation angle is 0.365 °, the pixel coordinate system is converted to the world coordinate system. After the edge segmentation based on curvature, the contour is fitted and the length is calculated. The total length of the contours is 38.00957 mm. This study first denoises the data, and then calculates the curvature. The edge of the contour is segmented and fitted automatically according to the curvature, and the contour is segmented again according to the curvature when the goodness of matching and the goodness of matching are different. The formulas for calculating the sweep angle and rotation angle of the elliptical arc are derived, and the indexes are calculated, and the automatic segmentation and fitting of the model are realized. Contour 1 is divided into 12 lines, 3 arcs and 17 elliptical arcs, and contour 2 is divided into 13 lines and 63 elliptical arcs. The highlight of this paper is that morphological denoising has obvious effect on burr edge and shadow, the gray moment sub-pixel method is used to extract edge contour with high precision, and the mapping relationship between pixel coordinates and world physical coordinates is realized by constructing rotation angle and proportion. The data are accurate and reasonable. The automatic cutting contour fitting method based on curvature is effective.