改进的基于凸壳仿射不变量的图像识别和配准算法  

Improved image recognition and registration algorithm based on convex hull and affine invariant

作　　者：朱政[1] 赵卫东[1] 王志成[1] 

机构地区：[1]同济大学CAD研究中心,上海200092

出　　处：《计算机应用》2007年第10期2559-2562,共4页journal of Computer Applications

摘　　要：提出了一种新的基于凸壳和仿射不变量的图像识别和配准的方法。该方法利用从参考图和测试图中得到的特征点提取其凸壳,并计算凸壳的仿射不变特征向量。通过比较参考图和测试图特征向量的一致性,建立它们的仿射变换关系,最后利用凸壳内特征点的匹配来实现识别和配准。该方法的优点是即使目标物体被部分覆盖或者缺损,其图像也能够达到较好的识别和配准效果。An improved image recognition and registration algorithm based on convex hull and affline invariant was presented. The convex hulls were extracted from the feature points of the reference and test images, and then the affine invariants were calculated. These invariants were used to establish the correspondence between the convex hull vertices of the test image and the reference image in order to undo the affine transformation between them, and then a feature points matching approach for recognition followed this. This method is very well suited to handle the occasions that the object in the image has undergone an affine transformation and is partly covered.

关 键 词：凸壳 仿射不变量 仿射变换 图像识别 图像配准 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]