论文部分内容阅读
对称是目标的基本形状特征,基于对称的形状表示和目标识别技术是模式识别和人工智能领域的重要研究方向。3-D空间中目标的对称表面在图像平面的投影往往呈现扭对称形态,利用扭对称信息可以加快诸如目标的姿态估计、方向计算以及图像校准等过程。本文详细地介绍了扭对称的形成模型、表示技术以及近来出现的三类主要的检测方法,即基于Hough变换的方法、基于不变特征的方法和基于规格化的方法,我们在叙述它们的主要思想、所适用的数据类型和优缺点后,给出了进一步研究的可能方向。
Symmetry is the basic shape feature of a target. Symmetric shape representation and object recognition are important research directions in pattern recognition and artificial intelligence. In the 3-D space, the projection of the target’s symmetry surface in the image plane often appears in the form of torsional symmetry. The use of the rotationally symmetric information can speed up the process of attitude estimation, direction calculation and image calibration. This paper introduces in detail the formation models of torsional symmetry, the representation techniques and the three main types of detection methods that have emerged recently, that is, Hough Transform-based methods, invariant feature-based methods and normalization-based methods. We describe their main Ideas, applicable data types and advantages and disadvantages, gives possible directions for further research.