地理科学 ›› 2015, Vol. 35 ›› Issue (6): 798-804.doi: 10.13249/j.cnki.sgs.2015.06.798

• • 上一篇    

ALOS 融合影像质量评价及其土地盐渍化应用研究

张利华1(), 翟靖超1, 李珊1, 杨金明2   

  1. 1.中国地质大学地球科学学院, 湖北 武汉 430074
    2.东北林业大学林学院, 黑龙江 哈尔滨150040
  • 收稿日期:2014-04-22 修回日期:2014-06-19 出版日期:2015-06-20 发布日期:2015-06-20
  • 作者简介:

    作者简介:张利华(1974-),女,河北邢台人,副教授,博士,主要从事流域环境演变分析和环境遥感研究。E-mail:huaz83@gmail.com

  • 基金资助:
    湖北省自然科学基金项目(2009CDB104)、中央高校基本科研业务费专项资金项目(2011019017)资助

Quality Evaluation and Land Salinization Classification Application on ALOS Image Fusion

Li-hua ZHANG1(), Jing-chao ZHAI1, Shan LI1, Jin-ming YANG2   

  1. 1.School of Earth Sciences, China University of Geosciences, Wuhan, Hubei 430074, China
    2. School of Forestry, Northeast Forestry University, Harbin, Heilongjiang 150040,China
  • Received:2014-04-22 Revised:2014-06-19 Online:2015-06-20 Published:2015-06-20

摘要:

将经过配准的同一地区不同空间分辨率和光谱分辨率的遥感影像进行融合是提高土地覆盖/土地利用分析精度的有效途径。采用PCA、IHS、HPF和小波变换融合法对内蒙古杭锦后旗中部地区的ALOS全色和多光谱影像进行融合,并对融合结果进行了定性和定量评价。基于地物光谱特征、解译标志和监督分类法提取试验区土地盐渍化信息,比较多光谱影像和融合影像的土地盐渍化信息提取精度。结果显示, PCA、IHS和HPF融合影像的空间细节表现能力得到提升,而PCA和小波变换融合影像的光谱保真度优于IHS和HPF融合影像;PCA融合影像的盐渍化分类精度、总分类精度和Kappa 系数均为最高,是最适于试验区土地盐渍化分类研究的融合方法。

关键词: ALOS, 影像融合, 土地盐渍化

Abstract:

Land salinization is a land degradation phenomenon which deteriorates the eco-environmental quality and agricultural production security, especially in arid and semi-arid areas. Particularly, the land salinization in Hetao irrigation area of Inner Mongolia (including Hangjinhouqi) is a major problem due to the arid climate, high salinity soil material, high mineralized groundwater, as well as high groundwater level caused by improper irrigation and drainage. Therefore, the monitoring of salinized land distribution is significant to prevent land salinization. Fused images based on different spatial and spectral resolutions are an important approach to improve the accuracy of land salinization classification. In this article, ALOS panchromatic and multi-spectral images of central Hanjinhouqi in Inner Mongolia, China, from August, 2010, were fused by employing the four image fusion methods, i.e., principal component analysis transform (PCA), intensity-hue-saturation transform (IHS), high pass filter (HPF) transform and wavelet transform. The effectiveness of each fusion method was evaluated qualitatively and quantitatively to examine the image quality and classification accuracy of land salinization. The result showed that: 1) spatial resolution of images improved after fused by PCA, IHS and HPF transform. 2) Image fused by HPF fused showed higher streaking noise. 3) Edge information of the object in wavelet transform image lowered compared to other fused image. 4) Spectral distortion of the images fused by PCA and wavelet transform was lesser than ones fused by IHS and HPF. In addition, the analysis of spectral signature showed that the mean gray value of different land cover pixel in the study area has the same change trend in the B2 and B3 bands, while different change trend was observed in B4 band because of the vegetation cover. The highest value of mean gray in the B2 and B3 bands was observed in resident cover, followed in sequence by salinized land, cultivated land, traffic land and water body. The highest value of mean gray in the B4 bands was observed in cultivated land .Furthermore, the land cover and land salinization information of researched area was also studied and extracted based on the interpreting marks, spectral signature and supervised classification. The extracted accuracy of multi-spectral images and fused images were compared as well. The classification results showed that the total classification accuracy and Kappa coefficient of PCA image, wavelet PCA image and wavelet IHS image are higher than multispectral images, while IHS image, HPF image and wavelet single band image are lower. The highest and the lowest value of total classification accuracy and Kappa coefficient were determined in PCA and HPF image respectively. The corresponding highest value of total classification accuracy and Kappa coefficient is 89.60% and 0.879 4 respectively while the corresponding lowest value is 65.20% and 0.654 2, respectively. Specifically, the PCA images had the highest classification accuracy of cultivated land (90.30%)and salinized land(90.90%) and HPF images had the lowest classification accuracy of cultivated land (69.23%)and salinized land(62.72%). The evaluation results of fused image quality and classification accuracy showed that PCA fused images is the best image for land use and land salinization information extraction in the study area.

Key words: ALOS, image fusion, soil salinization

中图分类号: 

  • P95