利用空间矩提取角特征

Corner Detection Using Spatial Moment

提出一种基于空间矩的特征角检测方法。首先基于Kitchen和Wang算法,提出梯度角变化和梯度幅值作为角特征检测规则;其次通过分析角特征模型,推导出空间矩多项式对角特征像素点的判别准则;最后利用非最大值压缩技术剔除虚假点,并进行准确定位。实验证明,空间矩角特征提取方法相对CSS和COP算法具有更好的准确性和鲁棒性。

A novel corner detection method based on spatial moments was presented in this paper. Firstly, based on the Kitchen's corner measure and the Wang's corner measure, gradient-direction change and image gradient are used as one decision rule of the corner detection. Then, by analyzing the corner-model, spatial-moment polynomials are employed as another decision rule of the corner detection. The corner feature can be detected by the two rules, but vertex of the corner can't be accurately localized. Therefore, non-max suppression method is used for reducing false points and improving localization. Experiment results illustrate that spatial moment corner detection approach has more accuracy and robustness than CSS and COP methods.