基于形态和约束结构的几何哈希法

几何哈希法,作为一种有效的模型搜索算法,在物体识别中有着重要的应用。现有的几何哈希法仅适合于仿射变换下的二维景物识别,论文提出了适合透视投影变换下三维物体识别的几何哈希方法。该方法利用物体的三维形态和物体中具有射影不变量的几何约束结构来构造哈希表。一方面,几何约束结构提供了物体模型的索引功能;另一方面,物体的三维形态提供了物体成像位姿的有关信息,使后续的匹配验证得以简化。实验中使用人造物体对该方法进行了验证,实验表明该方法正确有效。

Geometric Hashing Based on Aspects and Constrained Structures

Geometric hashing,as an effective model retrieving method,acts an important role in the object recognition. All the current geometric hashing methods are just suitable for the 2D object recognition under affine transformation.In this paper,a new geometric hashing method fitting for the 3D object recognition under perspective transformation is proposed.In which,3D aspects of object and geometric constrained structures possessing projective invariant are used to construct the hash table.In this way,geometric constrained structures provide the model retrieving function.On the other hand,the 3D aspects of object give the information of object pose,which can make the matching procedure simpler.In experiment,some artificial objects are used to verify the method and the experimental results show that the proposedmethod is correct and effective.