基于样本加权矩阵的自适应弱目标识别鲁棒算法

来源期刊:中南大学学报(自然科学版)2010年第3期

论文作者:陈炳权 刘宏立

文章页码:1052 - 1057

关键词:弱目标检测与跟踪;类矩阵;特征提取;自适应;鲁棒

Key words:weak target examination and track; kind of matrix; characteristics extraction; auto-adapt; robust

摘    要:以人脸为例,提出多区域特征融合的弱目标特征表示方法,并据此设计1种新的加权PCA-LDA(Principal element analysis-Linear distinction analysis)弱目标识别方法,用样本加权特征矩阵代替传统的特征向量,有效地将局部和全局特征加权融合。其步骤是:提取弱目标的全局特征,并用1种新的基于Gabor小波的方法提取其特征;基于将全局和局部特征加权融合的思想,得出权值的选择方法;通过研究多区域特征加权融合原理,提出定义每类样本的类矩阵,得到新的投影准则,解决小样本空间问题,最后得出加权PCA-LDA算法。研究结果表明:加权PCA-LDA算法能很好地结合弱目标图像全局和局部的互补信息,其识别效果优于各单一区域的分类效果。

Abstract: The weak target characteristic expression method of multi-provincial characteristics fusion was proposed based on human face. Accordingly, a new weighting PCA-LDA(Principal element analysis-Linear distinction analysis) weak target recognition method was designed, which replaces the traditional characteristic vector with the sample weighting eigen matrix and effectively fuses the partial and overall situation characteristic weighting. The procedures are as follows: The overall characteristics of weak target was withdrown based on the Gabor wavelet method. Next, the idea of overall and the partial characteristic weighting fusion were elaborated and the weight choosing method was given. Furthermore, the multi-provincial characteristics weighting fusion principle was elaborated, the method to define kind of matrix of each sample was proposed, and the new projection criterion was obtained. The little-sample space problem was solved, and the detailed weighting PCA-LDA algorithm was given. The experiment results show that the PCA-LDA algorithm can well unify the weak target image overall situation and the partial supplementary information, and the recognition effect of PCA-LDA algorithm surpasses that of each sole region.

基金信息:湖南省教育厅科学研究项目

相关论文

  • 暂无!

相关知识点

  • 暂无!

有色金属在线官网  |   会议  |   在线投稿  |   购买纸书  |   科技图书馆

中南大学出版社 技术支持 版权声明   电话:0731-88830515 88830516   传真:0731-88710482   Email:administrator@cnnmol.com

互联网出版许可证:(署)网出证(京)字第342号   京ICP备17050991号-6      京公网安备11010802042557号