| 摘 要: 当前,基于各向异性高斯方向导数的角点检测方法在处理对称特征时存在显著局限,导致检测精度不足。为此,提出一种改进的角点检测算法,该算法引入各向异性半高斯核函数,能够有效增强对称角点结构的特征响应,精准捕捉对称特征。同时,构建多尺度空间层级式验证机制,从多个尺度对检测结果进行验证,极大提升了检测的鲁棒性。在标准测试数据集上的实验结果显示,所提算法在旋转、缩放、剪切等6种变换类型下均展现出优异的稳定性,在视觉定位实验中定位精度达到94.1%。 |
| 关键词: 角点检测 半高斯核 各向异性高斯方向导数 对称性特征 层级式验证机制 |
|
中图分类号: TP391
文献标识码: A
|
|
| Corner Detection Based on Semi-Anisotropic Gaussian Directional Derivatives |
|
ZHANG Xinyue, LU Jin, WANG Haoyi, QIU Junjie, ZHANG Weichuan
|
(School of Electronic Information and Artificial Intelligence, Shaanxi University of Science and Technology, Xi’an 710021, China)
231612071@sust.edu.cn; lujin@sust.edu.cn; 231612067@sust.edu.cn; 2488602498@qq.com; zwc2003@163.com
|
| Abstract: Current corner detection methods based on anisotropic Gaussian directional derivatives exhibit significant limitations when handling symmetric features, leading to insufficient detection accuracy. To address this issue, this paper proposes an improved corner detection algorithm that introduces a sem-i anisotropic Gaussian kernel function. This approach effectively enhances the feature response of symmetric corner structures and accurately captures symmetric features. Additionally, a mult-i scale spatial hierarchical verification mechanism is constructed to validate detection results across multiple scales, significantly improving detection robustness. Experimental results on standard test datasets demonstrate excellent stability in affine transformation repeatability experiments under six types of transformations, including rotation, scaling, and shearing. In visual positioning experiments, the localization accuracy reaches 94.1% . |
| Keywords: corner detection sem-i Gaussian kernel anisotropic Gaussian directional derivatives symmetric features hierarchical verification mechanism |
