摘要
针对热红外遥感图像由于低对比度、条带噪声、低空间分辨率等特点而导致的检测效果不佳问题,提出了一种近岸舰船目标尺度自适应选择分层多阈值检测方法。采用舰船模板图像尺度归一化高斯拉普拉斯函数取极大值准则进行尺度自适应选择,利用所选的高斯多尺度空间差分多阈值筛选进行近岸舰船检测,并根据不同类型舰船模板图像尺度和分块数选择对热红外图像舰船目标检测的影响进行验证实验。实验结果表明:所提方法能根据模板尺度特征滤除相似区域,通过设置合理尺度和阈值参数能实现有效检测,且具有一定的抗噪能力。
In this paper, an approach to hierarchical multi-threshold detection of offshore ship targets based on scale-adaptive selection is proposed to address deficiencies in the recognition of TIR remote sensing images due to their characteristic low contrast,striping noise,and low spatial resolution.Scale-adaptive selection is conducted as per the maximum value of the normalized Laplacian of Gaussian for ship target image templates.Then offshore ships are detected through multi-threshold screening of the selected Gaussian multi-scale spatial differences. Finally, experiments are conducted to test the impact of scale selection as well as the blocking method for different types of ship image templates on the separation of ship targets in TIR images.The results show that the proposed approach can filter out similar regions according to template scales and achieve effective detection based on well-selected scale and threshold parameters, with a certain level of noise resistance.
出处
《海洋测绘》
CSCD
2017年第4期62-66,共5页
Hydrographic Surveying and Charting
关键词
热红外遥感
热红外图像
尺度选择
分层多阈值
舰船检测
目标检测
TIR remote sensing
TIR images
scale selection
hierarchical multi-threshold
ship detection
target detection
作者简介
马兰（1980-）,女,湖北南漳人,副教授,博士研究生,主要从事遥感图像识别研究.
