一种基于线性最小平方映射的目标分类算法  

An Object Classification Algorithm Based on Linear Least Squares Mapping Technique

作　　者：桑农[1] 魏智[1] 陈铭节[1] 彭嘉雄[1] 

机构地区：[1]华中理工大学图象识别与人工智能研究所

出　　处：《数据采集与处理》1996年第1期7-9,共3页Journal of Data Acquisition and Processing

摘　　要：介绍了一种基于线性最小平方映射的目标分类识别新方法。该方法先用自相关函数抽取目标图象特征，再用线性最小平方映射技术（ＬＬＳＭＴ）将所得特征向量映射到决策空间，在决策空间中用投影方法完成目标分类识别。作者用三类小汽车的二维二值图象进行实验，得到了比传统的Ｋ近邻、聚类等方法更好的分类结果。A new object classification algorithm based on linear least squares mapping technique (LLSMT) is presented. In the algorithm, we first extract the features of object image from autocorrelatiion function, then map these features into a decision space by using LLSMT, finally, classify the object by feature projection in the decision space. Experiments on 2D binary images of three classes of cars are performed.Results better than those from K-neighbour and other traditional methods are obtained.

关 键 词：目标分类 模式识别 图象识别 LISMT 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]