基于生物光学模型的水体多源遥感图像融合算法研究  

Research of Multi-Source Water Remote Image Fusion Algorithm Based on Bio-Optical Model

作　　者：郭宇龙[1] 王永波[1] 李云梅[1] 王桥[2] 朱利[2] 吕恒[1] 

机构地区：[1]南京师范大学江苏省碳氮循环过程与污染控制重点实验室,江苏南京210023 [2]环保部卫星环境应用中心,北京100029

出　　处：《光学学报》2015年第4期95-103,共9页Acta Optica Sinica

基　　金：国家自然科学基金(41271343);江苏高校优势学科建设工程项目(1411109012);高分辨率对地观测系统国家科技重大专项(05-Y30B02-9001-13/15-6)

摘　　要：针对内陆湖泊水环境遥感监测缺乏合适数据源这一问题,基于水体生物光学模型与传统图像融合算法,开发了一种适用于复杂内陆二类水体的生物光学融合(BOF)算法,用于融合多光谱数据和高光谱数据。利用Hyperion数据生成模拟数据集进行算法验证,并将实验结果与小波变换算法、Gram-Schmidt变换算法和色彩标准化算法分别进行对比,结果表明:从视觉效果来看,BOF算法较好地融合了高光谱数据的色彩信息和多光谱数据的空间细节信息;从图像精度指标来看,BOF算法不仅在多种分辨率差异下都得到最好的精度,且精度对分辨率差异不敏感;在叶绿素a浓度估算实验中,BOF算法也得到了最优的效果,均方根误差(RMSE)为9.817,其他三种算法的RMSE分别为18.841、15.913和15.655。新算法有较强的应用潜力,有望为内陆二类水体遥感监测提供更合适的数据源。Lacking of proper data source is becoming a problem for inland lake water environment remote sensing monitoring. To solve this problem, a bio- optical fusion（BOF） algorithm to merge multiband image and hyperspectral image of complex inland case 2 water is developed. The performance of the BOF algorithm is verified based on the bio-optical model of water body and traditional image fusion algorithm, by a Hyperion simulated dataset and compared with that of wavelet transform algorithm, Gram-Schmidt transforms algorithm, and color normalized algorithm. The results show that from the visual effect, the BOF algorithm has a better fusion of the color information of hyperspectral image and the detailed spatial information of the multispectral image. From the image evaluation indexes, the BOF algorithm gets the best performance under various resolution differences, and is not sensitive to the resolution difference; in the estimation experiment of chlorophyll- a concentration, the BOF algorithm also gets the best result, the root mean square deviation（RMSE） is 9.817, meanwhile, RMSE of the other three algorithms are 18.841,15.913 and 15.655. The proposed algorithm has strong potential applications, and expected to provide better data source for inland case 2 water remote sensing monitoring.

关 键 词：遥感 图像融合 生物光学模型 内陆二类水体 HYPERION 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]