基于HSI高光谱和TM图像的土地盐渍化信息提取方法

选择黄河三角洲垦利县代表性盐碱化区域为研究区，以2011年3月15日HJ-1A卫星HSI高光谱影像和2011年3月22日TM影像为信息源，经几何纠正、图像裁剪、大气校正等预处理，分析不同盐渍化程度土地、水体、滩涂等主要地类的光谱特征，确定地类信息提取特征波段。结合土壤盐分含量，采用定量与定性相结合规则，构建地类信息提取模型，以决策树分类方法进行图像分类，提取土地盐渍化信息。利用地表点位土壤含盐量数据对地表土地盐渍化程度的化学分析结果，对遥感解译数据进行精度验证，并对高光谱和多光谱影像的分类精度进行比较分析。结果表明：HSI图像的总体分类精度达96.43%，Kappa系数为95.59%，而TM图像的总体分类精度为89.17%，Kappa系数为86.74%，说明相比多光谱TM数据，基于高光谱图像可以更为准确有效地提取土地盐渍化信息;由分类结果图可以看出，高光谱影像土地盐渍化的区分度高于多光谱影像。该研究探索了高光谱图像土地盐渍化信息的提取技术方法，提供了不同盐渍化土地的分布比例数据，可为黄河三角洲滨海盐碱土地资源的科学利用与管理提供决策依据。

Land Salinization Information Extraction Method Based on HSI Hyperspectral and TM Imagery

This paper chose the typical salinization area in Kenli County of the Yellow River Delta as the study area, selected HJ-1A satellite HSI image at March 15, 2011 and TM image at March 22, 2011 as source of information, and pre-processed these data by image cropping, geometric correction and atmospheric correction. Spectral characteristics of main land use types including different degree of salinization lands, water and shoals were analyzed to find distinct bands for information extraction. Land use information extraction model was built by adopting the quantitative and qualitative rules combining the spectral characteristics and the content of soil salinity. Land salinization information was extracted via image classification using decision tree method. The remote sensing image interpretation accuracy was verified by land salinization degree, which was determined through soil salinity chemical analysis of soil sampling points. In addition, classification accuracy between the hyperspectral and multi-spectral images were analyzed and compared. The results showed that the overall image classification accuracy of HSI was 96.43%, Kappa coefficient was 95.59%; while the overall image classification accuracy of TM was 89.17%, Kappa coefficient was 86.74%. Therefore, compared to multi-spectral TM data, the hyperspectral imagery could be more accurate and efficient for land salinization information extraction. Also, the classification map showed that the soil salinity distinction degree of hyperspectral image was higher than that of multi-spectral image. This study explored the land salinization information extraction techniques from hyperspectral imagery, extracted the spatial distribution and area ratio information of different degree of salinization land, and provided decision-making basis for the scientific utilization and management of coastal salinization land resources in the Yellow River Delta.