作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2008, Vol. 34 ›› Issue (15): 19-21. doi: 10.3969/j.issn.1000-3428.2008.15.007

• 博士论文 • 上一篇    下一篇

星载SAR图像机场兴趣区检测算法

张绍明1,陈映鹰1,林 怡1,胡希驰2   

  1. (1. 同济大学遥感与空间信息技术研究中心,上海 200092;2. 中国科学院电子学研究所,北京 100080)
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2008-08-05 发布日期:2008-08-05

Detection Algorithm for Airport ROI in Spaceborne SAR Image

ZHANG Shao-ming1, CHEN Ying-ying1, LIN Yi1, HU Xi-chi2   

  1. (1. Research Center of Remote Sensing and Spatial Information Technology, Tongji University, Shanghai 200092; 2. Research Institute of Electronics, Chinese Academy of Sciences, Beijing 100080)
  • Received:1900-01-01 Revised:1900-01-01 Online:2008-08-05 Published:2008-08-05

摘要: 提出一种解决低信噪比、大尺寸的星载合成孔径雷达(SAR)图像机场兴趣区(ROI)自动检测问题的方法。对图像进行基于马尔可夫场的分割,由目标标记方法确定ROI的候选区。用Hough变换检测跑道平行线,滤除部分伪ROI。计算ROI候选区的跑道方向投影直方图及目标的宽、高、宽高比、目标背景面积比和目标背景灰度比5个参数,用支持向量机对样本参数进行学习,完成ROI的最终判定。用实际星载SAR图像进行试验,结果表明了该方法的有效性和可靠性。

关键词: 合成孔径雷达图像, 机场兴趣区, 马尔可夫随机场, 投影直方图, 支持向量机

Abstract: This paper proposes a novel airport Region of Interest(ROI) detection algorithm for large-scale spaceborne Synthetic Aperture Radar(SAR) image. The image is segmented via ICM algorithm according to the frame of MAP and the model of markov random field. The candidate ROIs are recorded by using object labeled algorithm. The parallel is detected by Hough transformation and the ones without parallel are removed. The histogram of projection of the candidate ROIs are calculated and the sample vector of candidate ROIs are constructed by five factors, including width and height of object, Ratio between Width and Height(RWH), Area Ratio between Object and Background(AROB) and Gray Mean between Object and Background(GMROB). Support Vector Machine(SVM) classifier is trained and applied to give the final result of ROI detection. Experiment with actual data indicates the algorithm is effective.

Key words: Synthetic Aperture Radar(SAR) image, airport Region of Interest(ROI), Markov random field, projection histogram, Support Vector Machine(SVM)

中图分类号: