基于滑动窗口的自适应角点检测研究

Adaptive corner detection based on slide windows

基于边界曲线的角点检测算法存在离散化误差较大、角点定位精度差等缺陷,本文提出了基于曲率的自适应角点检测算法,设计了滑动窗口自适应策略,并用于边界点局部支持邻域的选择,克服了以往局部支持域不变的缺陷;在滑动窗口的基础上提出了基于特征向量夹角的曲率计算方法,从而根据曲率值进行角点的提取。本角点检测算法能对角点进行精确定位,易于实现。仿真结果验证了本算法的可行性与优越性。

Boundary-based corner detection algorithm has many shortcomings, such as the error of discretization and poor localization accuracy of corner point. Based on curvature and slide windows, an adaptive corner detection algorithm is proposed. Adaptive slide window strategy is applied to select the optimal support field of boundary point. The new curvature calculation method is presented to transform two-dimension planar curve to one-dimension curve, then the singularities of one-dimension curve means the corner point on planar curve. The simulation results indicate the superiority of our algorithm to these other methods.