光学学报, 2020, 40 (17): 1728001, 网络出版: 2020-08-25
四波段多光谱遥感图像的船舶目标显著性检测 下载: 924次
Saliency Detection for Ship Targets on Four-Band Multi-Spectral Remote Sensing Images
遥感 图像处理 多光谱遥感 显著性检测 船舶检测 remote sensing image processing multispectral remote sensing saliency detection ship detection
摘要
针对复杂背景下船舶目标检测率低和光学遥感传感器的多光谱近红外(NIR)波段利用率低的问题,提出一种四波段多光谱遥感图像船舶目标显著性检测算法。所提算法利用四波段遥感数据中可见光波段图像的色彩内容饱满、NIR图像细节突出的特点,首先将可见光蓝、绿、红三通道图像变换到CIE-Lab色彩空间;然后对NIR图像进行非下采样轮廓波变换分解,对得到的高频分量进行非线性增强,以抑制噪声并增强细节,对低频分量进行反锐化掩模处理增强,以改善图像亮度的均匀性,并将高频分量和低频分量与Lab空间的亮度图像相结合,得到新的Lab图像;最后利用最大对称环绕模型对Lab图像进行显著性分析,得到船舶目标的显著性图像。实验结果表明,所提算法能够充分抑制云层、海浪尾迹等杂波干扰的复杂背景信息,同时在低对比度背景下能够突出船舶目标,具有高的查准率和查全率。
Abstract
To solve the problems of low ship-target detection rate and low multi-spectral near-infrared (NIR) band utilization rate of optical remote sensors in complex background, a novel algorithm for saliency detection of ship targets based on four-band multi-spectral remote sensing images is proposed. The proposed algorithm employs the features of visual images in four-band remote sensing data that have rich color information as well as NIR images that have good ability to describe details. First, the three bands of blue, green, and red images are transformed into the CIE-Lab color space. Then, the NIR image is decomposed via the non-subsampled contourlet transform. The obtained high-frequency components are nonlinearly enhanced to suppress noise and enhance details, and the low-frequency components are enhanced via unsharp masking to improve the uniformity of image brightness. The high-frequency components and low-frequency components are combined with the brightness images in Lab space to obtain a new Lab image. Finally, the maximum symmetric surround model is applied to the new Lab image to obtain a saliency image of the ship target. The experimental results show that the proposed algorithm can fully suppress the complex background information of clutter interferences, such as cloud waves and sea wakes, and it also can highlight ship targets in low contrast backgrounds. The proposed algorithm has good precision and recall.
王文胜, 黄民, 李天剑, 胡欢, 毕国玲. 四波段多光谱遥感图像的船舶目标显著性检测[J]. 光学学报, 2020, 40(17): 1728001. Wensheng Wang, Min Huang, Tianjian Li, Huan Hu, Guoling Bi. Saliency Detection for Ship Targets on Four-Band Multi-Spectral Remote Sensing Images[J]. Acta Optica Sinica, 2020, 40(17): 1728001.