摘要
本文提出了一种快速有效的空间平面曲线的识别方法。在这种方法中,从单幅图像提取表征曲线形状的描述子,该描述子由曲线上透视不变点之间的射影距离不变量构成,且不受物体和摄象机之间的位置、方向以及摄象机物理参数的影响,因此这种识别方法不受图象透视的限制。本文用一组真实的平面物体进行了识别实验,结果表明这些形状描述子对空间平面曲线的识别有很好的性能。
An effective method for recognition of planar curves is presented in the paper. In the method shape descriptors are extracted from single view image and constructed by projective distances between perspectively invariant points on the curves. The descriptors are unaffected by the position and orientation of the curves with respect to the camera, and the intrinsic parameters of the camera. As a result, the recognition method is immune from all the perspective effects. The shape descriptors were tested on recognition of planar curve objects. The results of the tests show that the descriptors developed in the paper have very powerful performances for recognition of planar curves.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
1995年第2期108-113,共6页
Pattern Recognition and Artificial Intelligence
基金
国家自然科学基金(69275004)
关键词
平面曲线
识别
机器视觉
透视不变量
Perspective Invariant, Projective Distance, Shape Description, Curve Recognition.
