摘要： 提出了局部特征和全局特征相结合的角点无阈值识别的思想和算法，同时提出了4 个重要的概念及其算法：边界局部角度、相似度、高度、线性度，作为实现的基础。角度和相似度描述角点的局部特征；高度和线性度描述角点的全局特征，4 者的拟合函数为F。F变化率的最大值点自动确定角点和非角点的分界。实验结果证实了该算法角点定位的准确性、旋转不变性，具有一定的尺度不变性和抗噪声能力。

关键词: 相似度；线性度；角点局部特征；角点全局特征；无阈值；角点探测；逼近多边形

Abstract: This paper proposes a new corner detector idea and algorithm without threshold which utilities the local and global features of corners.Four notions, which are local angle of boundary, similarity, altitude, linearity, and its calculation are proposed as the base of the corner detector.Angle and similarity describe the local features of corners; altitude and linearity describe the global features of corners. F is the function of the fourfeatures. The maximum ratio of F is able to cut a clear line between corners and non-corners automatically. Experiments are performed to show thatthe proposed algorithm can locate corners accurately, is rotationally invariant, has scale invariant property and is insensitive to noise contamination.

Key words: Similarity; Linearity; Local feature of corner; Global feature of corner; Without threshold; Corner detection; Close polygon