电子学报 ›› 2020, Vol. 48 ›› Issue (3): 431-441.DOI: 10.3969/j.issn.0372-2112.2020.03.003

• 学术论文 • 上一篇    下一篇

基于trace特征的ISAR像空间目标识别

杨虹1, 张雅声2, 徐灿2   

  1. 1. 航天工程大学研究生院, 北京 101416;
    2. 航天工程大学, 北京 101416
  • 收稿日期:2019-01-06 修回日期:2019-11-25 出版日期:2020-03-25
    • 作者简介:
    • 杨虹 女,1991年出生,四川省绵竹人.现为航天工程大学宇航科学与技术系博士研究生,研究方位为基于逆合成孔径雷达图像的空间目标识别.E-mail:1558513572@qq.com;张雅声 女,1974年出生,安徽省淮南人.航天工程大学博士,教授,博士生导师.主要从事航天任务分析与设计,星座设计等方面的研究.E-mail:13521219203@139.com
    • 基金资助:
    • 国家自然科学基金 (No.61304228)

Space Target Recognition Based on Trace Feature of ISAR Image

YANG Hong1, ZHANG Ya-sheng2, XU Can2   

  1. 1. Graduate School, Space Engineering University, Beijing 101416, China;
    2. Space Engineering University, Beijing 101416, China
  • Received:2019-01-06 Revised:2019-11-25 Online:2020-03-25 Published:2020-03-25
    • Supported by:
    • National Natural Science Foundation of China (No.61304228)

摘要: 论文提出了一种基于trace特征的逆合成孔径雷达(Inverse Synthetic Aperture Radar,ISAR)像空间目标识别算法.首先将ISAR像进行分割与归一化处理,利用Canny边缘检测、Hough变换方法提取空间目标ISAR像最长轴,确保所提特征具有旋转不变性;然后仅对最长轴所在局部区域进行Trace变换生成空间目标ISAR像的局部trace矩阵,使得所提trace特征满足低维要求;再将trace矩阵每一列向量进行移位对准操作以消除ISAR像平移对识别带来的影响并将其作为空间目标识别的特征向量;最后在特征空间内以最小欧氏距离作为不相似度,采用集成分类器AdaBoost.M2-KNN完成了5类空间目标的分类识别.通过5类空间目标的ISAR数据对该方法进行目标识别验证,并与现有的几种ISAR像特征提取方法进行了对比.结果表明论文所提算法可行有效,可以明显地提高识别率.

关键词: Trace变换, ISAR像, 集成分类器AdaBoost.M2-KNN

Abstract: This paper proposes a space target recognition algorithm based on the trace feature of ISAR (Inverse Synthetic Aperture Radar) image. Firstly, the ISAR images are segmented and normalized. The Canny edge detection and Hough transform method are used to extract the longest axis of the ISAR image of space target to ensure the rotation invariance of the proposed feature. Then, Trace transformation is only carried out in the local area where the longest axis is located to generate the local trace matrix of ISAR image, so that the trace features proposed can meet the requirements of low dimension; then each column of the trace matrix is shifted and aligned to eliminate the impact of ISAR image translation on the recognition and take it as the feature vector of the space target recognition; finally, taking the minimum Euclidean distance as the dissimilarity in the feature space, AdaBoost.M2-KNN, an ensemble classifier, is used to classify and recognize the five types of space objects. The method is validated by the ISAR data of five types of space targets, and compared with several existing ISAR image feature extraction methods. The results show that the proposed algorithm is feasible and effective, and can significantly improve the recognition rate.

Key words: trace transform, ISAR image, ensemble classifier AdaBoost.M2-KNN

中图分类号: