基于高光谱图像的未知海域水下礁滩探测方法研究

Underwater reefs are very important for human activities on the sea. There are many underwater reefs in East China Sea and South China Sea, and the requirement of detecting underwater reef is more and more imperious. Current studies on hyperspectral images of underwater reefs are bathymetry retrieval (including its application for underwater object detection) and coral reef monitoring. However, these methods can not be used for the detection of underwater reefs in unknown sea, since they depend on the bathymetry reference points. This project plans to do the following studies: 1) Analyze the influence of second-order light and other effects, then develop a spectral correction method for hyperspectral images of underwater reefs; 2) Develop the spectral feature extraction method based on spectral derivatives, then develop the band selection methods to select the bands where underwater reef disappears, and develop extraction methods for the image change features of underwater reef; 3) Based on the one-dimensional spectral and two-dimensional image change features of underwater reef, develop the fusion methods to detect underwater reef in the feature and decision level. The proposed methods can be useful for detecting the underwater reef around the Nansha (Spratly) Islands.

暗礁和暗滩等水下礁滩对人类海上活动具有重要意义。中国东海和南海有许多水下礁滩，探明上述水下礁滩情况的需求也越来越迫切。现有水下礁滩高光谱研究主要为水深反演（包括在水下目标检测的应用）和珊瑚礁监测，但相应检测方法依赖于水深参考点，并不适合未知海域的水下礁滩探测。本项目拟展开的具体研究包括：1）对二阶光效应等影响进行分析，研究光谱校正方法；2）基于光谱导数法研究水下礁滩光谱特征提取方法，研究波段选择方法用于选择水下礁滩在近红外波长附近逐步消失的波段，研究水下礁滩图像变化特征提取方法；3）基于一维光谱特征和二维图像变化特征，研究特征层和决策层的融合探测方法。项目成果能够应用于探测南沙岛礁等周边水下礁滩。

基于高光谱图像的未知海域水下礁滩探测方法研究从2015年1月开始，为期三年。针对申请书中提出的水下礁滩探测的研究思路开展了深入的研究，在水下礁滩光谱校正与去噪，水下礁滩光谱和图像特征分析以及水下礁滩融合探测进行了重点研究。主要成果包括：（1）提出了基于低秩结构的噪声估计方法；（2）提出了引入光谱导数的水下礁滩光谱特征提取方法；（3）提出了一种基于图像差值和比值的水下礁滩图像特征提取的方法；（4）提出了基于Ln比值的水下礁滩特征提取方法；（5）提出了一种基于堆栈泛化的光谱和图像融合探测方法。.在算法研究的基础上，建立了拥有全部知识产权的水下探测方法。从获得批准日起至2017年12月共发表论文5篇，其中SCI检索1篇，EI检索3篇，申请国家发明专利3项，授权1项。