基于多尺度特征聚类算法的不确定目标检测  

Indeterminate Target Detection Based on Multi-scale Feature Clustering Algorithm

作　　者：周颖[1] 赵晓哲[1] 逯超 ZHOU Ying;ZHAO Xiao-zhe;LU Chao(School of Electronics and Information,Northwestern Polytechnical University,Xi’an 710072,China)

机构地区：[1]西北工业大学电子信息学院,西安710072

出　　处：《火力与指挥控制》2019年第4期164-168,共5页Fire Control & Command Control

摘　　要：提出一种基于无样本的SAR图像目标检测分类方法。针对空战过程中难以获得大量SAR图像目标样本问题,采用基于全局CFAR的多尺度SIFT特征进行目标纹理描述,并针对特征维度较高及特征描述之间存在的冗余问题,通过PCA算法对其进行降维处理后,采用DBSCAN算法对潜在目标区域的多尺度SIFT特征进行分类实现目标检测。通过单一目标和多类目标图像进行实验验证,实验结果表明该方法具有一定的有效性和可行性。A sample-free SAR image target detect classification method is proposed in this thesis.Aiming at the problem that it’s difficult to obtain plenty of SAR image objective samples in the process of air combat,the multi-scale SIFT feature based on global CFAR is adopted to describe the objective texture,and the dimensionality reduction processing is applied by PCA algorithm for the problem of higher feature dimensions and the redundancy between feature descriptions.DBSCAN algorithm is used to classify the multi-scale SIFT features of the potential target region to achieve the objective detection in final. The single target and multiple objective images are used to verify the feasibility and effectiveness of the proposed method in the end.

关 键 词：SAR图像 SIFT特征 聚类算法 目标分类 

分 类 号：TN957.52[电子电信—信号与信息处理]