西安电子科技大学学报

• 研究论文 • 上一篇    下一篇

场景语义SAR图像桥梁检测算法

黄勇1,2;刘芳1,2   

  1. (1. 西安电子科技大学 计算机学院,陕西 西安 710071;
    2. 西安电子科技大学 智能感知与图像理解教育部重点实验室,陕西 西安 710071)
  • 收稿日期:2017-08-24 出版日期:2018-08-20 发布日期:2018-09-25
  • 作者简介:黄勇(1974-),男,西安电子科技大学博士研究生,E-mail:chinahe0609@aliyun.com
  • 基金资助:

    国家重点基础研究发展计划资助项目(2013CB329402);国家自然科学基金资助项目(61573267);重大研究计划资助项目(91438201, 91438103)

Detecting water bridge in SAR images via a scene semantic algorithm

HUANG Yong1,2;LIU Fang1,2   

  1. (1. School of Computer Science and Technology, Xidian Univ., Xian 710071, China;
    2. Ministry of Education Key Lab. of Intelligent Perception and Image Understanding, Xidian Univ., Xian 710071, China)
  • Received:2017-08-24 Online:2018-08-20 Published:2018-09-25

摘要:

在合成孔径雷达图像中存在大量相干斑乘性噪声,导致桥梁检测中极易出现漏检和误检.针对这一问题提出了基于场景语义的水上桥梁检测算法.该方法首先根据图像的内容自动分割出水域和陆地两类不同的场景,沿着水陆交界区域搜索疑似桥梁目标,有效地缩小了搜索范围;然后提取图像的素描特征,抑制相干斑噪声的干扰;最后根据素描特征的几何特性定义桥梁的隶属度函数,实现桥梁目标的检测.在真实合成孔径雷达图像上的仿真实验表明,该算法有效地降低了桥梁的漏检率和误检率,具有较好的鲁棒性.

关键词: 图像处理, 目标检测, 桥梁检测, 合成孔径雷达图像, 场景语义

Abstract:

Much speckle noise causes the bridge to be missing or undetected in SAR images. Aiming at the problem, we propose a novel algorithm based on scene semantic water bridge recognition. First, the image is automatically segmented into land and water scenes which narrow the search region of the bridge to the junction area. Second, the Primal Sketch features of the image are extracted to suppress the interference of speckle noise. Last, the membership function of the bridge is defined by its geometric characteristics based on the Prime Sketch features. Experiments show that the proposed algorithm effectively reduces the missing rate and error detection rate of the bridge, and has better robustness.

Key words: image processing, target detection, bridge detection, synthetic aperture radar image, scene semantic