摘要
针对高分辨遥感图像中目标背景为复杂多变的海面,提出一种基于邻域特性分析的海面舰船检测方法.根据邻域窗口的均值方差积特性,初分割消除大部分海面背景;再通过后续邻域均值滤波消减海面分散杂波,利用形态学处理进一步消除杂波干扰,确定目标候选区域;最后结合舰船相关特征(长宽比和矩形度等)对目标进行验证,排除虚警,输出最终检测结果.经对大量不同海面背景进行验证表明,该方法适应性强,稳定性好,检测率高.
A ship deteclion method is proposed based on the analysis of neighborhood features for high-resolulion sea surface remote sensing images with complex background. According to the variance of neighborhood window, the ini- tial segmentation eliminates most of the background of the sea surface, and then a filter of neighborhood average is used to subtrac｜ the dispersion of sea cluner. Afterward, morphological processing is utilized to fimhcr eliminale clutter and determine the target candidate region. Finally, ship characteristics （aspect ratio, rectangular degree, etc. ） are combined to verify the target detected, eliminate false alarms, and output the final tesl results. Tcslcd through a large number of experiments with different sea backgrounds, the results show that the proposed method is of good flexibility, stability, and detection rate.
出处
《深圳大学学报（理工版）》
EI
CAS
北大核心
2013年第6期584-591,共8页
Journal of Shenzhen University(Science and Engineering)
基金
国家自然科学基金资助项目(61071206)
国防预研基金资助项目(9140***9302)~~
关键词
海洋遥感图像
舰船检测
邻域特性分析
阈值分割
形态学处理
目标判决
ocean remote sensing image
ship detection
neighborhood characteristic： analysis
threshold segmenta-tion
morphological processing
target judgment
作者简介
龚志成（1988-）,男（汉族）,湖南省常德市人,深圳大学硕士研究生.E-mail：riky_cheng@qq.com
