基于距离信息扩展原理的物体识别方法  

Using Range Information Propagation Transform in Object Recognition

作　　者：陈向荣[1] 朱志刚[1] 石定机[1] 

机构地区：[1]清华大学计算机科学与技术系

出　　处：《中国图象图形学报（A辑）》1997年第8期614-618,共5页Journal of Image and Graphics

摘　　要：利用稀疏距离图象与灰度图象相结合，通过射影变换可使灰度图象特征有条件地获取距离信息，这称为基于射影变换的距离信息扩展原理。利用这些距离信息可计算物体表面平面图形的欧氏空间不变量并对景物中的物体产生假设。由于初始假设建立在可靠的基础上，并使用了带反馈的混合控制策略，使这种基于模型的物体识别定位技术具有实用的前景。此文讨论了整个方法的原理及其实施方案，并实验验证了原理的正确性与有效性。s Using sparse range image and gray level image in combination, features extracted from the gray level image can be assigned conditionally with corresponding range values by means of a projective mapping. It is called range information propagation transformation. The range information propagated in this way can be used to calculate the Euclidean invariant of corresponding surface planar patterns and hence to generate initial hypothesis. Since initial hypotheses are extracted based on reliable information, and hybrid control strategy with feedback is used, the proposed method can be used in practice situations. In this paper the principle of the method and its implemental issues are addressed, and some preliminary experimental results are presented to demonstrate its effectiveness and efficiency.

关 键 词：物体识别 数据融合 模式识别 距离信息扩展 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]