基于邻域分析的海洋遥感图像舰船检测方法  

A method for ship detection based on neighborhood characteristics in remote sensing image

作　　者：龚志成[1] 曾惠翼 裴继红[1] 

机构地区：[1]深圳大学ATR国防科技重点实验室,深圳518060

出　　处：《深圳大学学报（理工版）》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)~~

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.

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 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]