Edge Detection of Depth Image Based on Contour Curve

2019 International Conference on Advanced Electronic Materials, Computers and Materials Engineering

无线电电子学;计算机科学;材料科学

Wu, Yuzhe^1

School of Information Science and Technology, Agricultural University of Hebe, Baoding

710000, China^1

关键词: Arbitrary points;    Candidate list;    Degree of approximation;    First principal components;    Important features;    Nonlinear extension;    Recognition and locations;    Recognition efficiency;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/563/5/052037/pdf
DOI  :  10.1088/1757-899X/563/5/052037

来源: IOP

【 摘 要 】

Edge is one of the most important features of an image because it recognizes the world through images and understands the position and contour of objects in images through edges. Edge monitoring of depth image is the basis of object recognition and location. The contour curve calculates the feature points and maps them to the original contour space. The original contour is matched by a sequence of markers. At the initial stage of recognition, the dissimilar targets and different gestures are eliminated quickly. The candidate list of targets is generated for accurate matching to improve the recognition efficiency. The curve is a nonlinear extension of the first principal component, emphasizing the search for the "middle" through the distribution of data, starting from an arbitrary point of the image, estimating the pixel points near the point and obtaining the gradient value. If the point satisfies the characteristics of the edge point, then the point detour depth traverses to find edge points that are perpendicular to the direction of the gradient and marks the visited points. And satisfy the self-consistent smooth curve without parameters, these curves give an overview of the data. Determine the degree of approximation; when the judgment has converged to the target contour, continue to converge and penetrate the concave area of the target contour.