仿射坐标系下多视点的三维物体识别方法研究

Study on Approach of 3D Object Recognition Across Multiple Views in Affine Coordinate System

多个视角下三维物体的识别一直是计算机视觉和模式识别领域的重要研究方向。对平面区域或适当的变换区域中以不同视角观察时图像中的点集所具备的性质进行了研究,并得出了在仿射变换下点模式匹配所具备的重要性质,即基于不同视角下平面图形的傅立叶描绘矩阵,得到形状匹配的数学约束,并将其应用于多个视角下物体的形状识别。理论分析及仿真实验表明,该方法在形状识别中是简便、快速、有效的。

3D object recognition across multiple views is an important problem in the flied of computer vision and pattern recognition. This paper payed attention to the properties of a collection of primitives in multiple views,and derived important properties for point-pattern matching under affine transformation ,which based on the rank of a matrix of Fourier domain descriptor. Algebraic was derived constraint for shape recognition ,and was use to shape recognition.