The comprehensive and recognition of 3D scene based on invariants are the most important research areas in computer vision fields. The conventional studies of invariants are that these invariants are derived for planar objects using points, lines, and conics from images. The invariants of triangular, quadrilateral, pentagonal and hexagon are structured by use of triangular area from 2D points perspective projection in this paper. And the computational methods of the invariants are derived. The result of example shows that this formula is correct on the basis of studies.%不变量的场景理解和目标识别是计算机视觉研究的一个重要领域,以往有关不变量研究主要集中在点、直线、二次曲线等几何元素之间。在二维平面点的射影变换的基础上,利用平面三角形面积不变量构造了三角形、四边形、五边形、六边形等共面多边形的不变量,并提出了具体的计算方法。在此基础上通过举例分析和实验验证,证明文中所给公式的正确性。
展开▼