基于色彩空间变换与异常检测的多光谱舰船感兴趣区域检测  

Multispectral ship ROI detection based on color space transformation and anomaly detection

作　　者：唐天翼[1] 朱长仁[1] 赵和鹏[2] 

机构地区：[1]国防科学技术大学电子科学与工程学院ATR重点实验室,湖南长沙410073 [2]海装电子部,北京100841

出　　处：《现代电子技术》2015年第21期28-31,35,共5页Modern Electronics Technique

摘　　要：针对多光谱遥感图像中复杂海面背景下海上舰船检测问题,在高斯模型RX异常检测算法的基础上进行改进,提出一种基于色彩空间变换的复杂海面上舰船感兴趣区域(ROI)的自动检测方法,解决了传统的阈值分割方法在遥感图像复杂海面背景下较难将目标与背景分离的问题。在多种不同复杂海面背景下的舰船ROI检测实验中,与传统阈值分割方法比较,该舰船ROI检测方法有较好的检测效果。Since it is difficult to detect the ships with complex sea background in multispectral remote sensing images,with the improvement on the basis of RX（Reed-Xiaoli）anomaly detection algorithm of Gaussian model,an automatic detection method of region of interest（ROI）of ship in complex sea based on color space transformation is proposed. The problem that the traditional threshold segmentation methods are hard to separate the target in the remote sensing image from the complex sea was solved. Compared with the traditional threshold segmentation methods,the ship ROI detection method has better detection results in ship ROI detection experiments under various complex sea background.

关 键 词：多光谱遥感 色彩空间变换 RX算法 舰船感兴趣区域检测 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置] TP391[自动化与计算机技术—控制科学与工程]