摘要
目标物体的识别和匹配在计算机视觉、图像视频压缩与传输中都有重要应用 .由于隐含多项式曲线对物体有良好的描述能力 ,因而用它识别和匹配目标物体是比较有效的 .文章首先证明了隐含多项式曲线封闭有界的充要条件定理 ,接着基于隐含多项式曲线的二次分解性质 ,给出了目标物体仿射几何不变量的计算方法 .实验证明这种基于首二次因子积的仿射几何不变量准确的描述了物体的特征 ,从而能较好的设别出复杂的 。
Objects recognition and match have important applications in computer vision, image video compression and transmission. Due to the advantages of the describing objects, implicit polynomial curves are efficient to recognize objects. First, this paper proves the necessary and sufficient condition of existence of the bounded and closed IP curves. Then, the paper presents a method for calculating affine geometric invariants of objects based on decomposed conics of the implicit polynomials curves. The experiment proves that affine geometric invariants are robust and efficient to recognize complicated objects, even to recognize objects with some information lost.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2004年第12期1987-1991,共5页
Acta Electronica Sinica
基金
国家 8 63项目 (No .2 0 0 2AA42 32 0 0 )
关键词
隐含多项式曲线
仿射几何不变量
目标识别
计算机视觉
Computational geometry
Curve fitting
Image communication systems
Image compression
Object recognition
Polynomials
Theorem proving
