图像局部矢量点结构的角点检测算法

Corner Detection Algorithm for Image Local Vector Point Structure

针对基于边缘的角点检测算法只关注角点位置信息而忽略局部信息的问题,以及去除伪角点和减少角点遗漏现象提高角点检测性能,提出一种基于矢量点结构的角点检测算法。首先提取图像边缘轮廓,在图像边缘基础上提出一种边缘矢量点结构,在边缘像素点邻域提取带有方向的特征点,聚类特征点形成邻接图像边缘的链,融合同一区域相邻链的相邻端点形成最终角点。该算法在多种场景下与ORB算法、SIFT算法以及SURF算法相比,提出的算法获取的角点定位性优于其他3种算法。从图像边缘轮廓着手,提取位于角内侧区域的角点,屏蔽背景影响,以提高角点检测器的检测性能。实验结果表明,该算法拥有良好的角点检测稳定性与定位性能。

To solve the problem that corner detection algorithm based on edge only focuses on the location information of corners and ignores local information, and to improve corner detection performance by removing pseudo-corners and reducing corner omissions, proposes a corner detection algorithm based on vector point structure. Firstly, extracts the edge contour of the image, and then proposes an edge vector pointstructure based on the edge of the image. Extracts the directional feature points from the neighborhood of the edge pixels, and clusters thefeature points to form a chain of adjacent edges of the image. Compared with Harris algorithm, ORB algorithm and SURF algorithm, the proposed algorithm is superior to the other three algorithms in many scenarios. Starting from the edge contour of the image, corner points locat-ed in the medial corner region are extracted to shield the background effect, so as to improve the detection performance of corner detector.Experimental results show that the algorithm has good corner detection stability and location performance.