论文部分内容阅读
土壤盐渍化是制约农业生产和发展的主要障碍。目前土壤盐渍化的遥感监测主要基于中、低分辨率卫星影像,并采用传统的基于像元分类方法,对盐渍化信息的细节描述不足,监测精度不高。本文使用国产GF-1影像,结合自上而下的多尺度分割技术和面向对象的信息提取技术,针对田间尺度下的盐渍化信息进行精确地提取、分类,并与传统分类方法进行了对比。结果表明:面向对象法和最大似然法的分类总体精度分别为92.72%和84.31%,Kappa系数分别为0.90和0.78。该技术能准确提取田间尺度下的盐渍地信息,在未来的农田盐渍化高精度监测研究中具有一定应用价值和发展潜力。
Soil salinization is a major obstacle to agricultural production and development. At present, the remote sensing monitoring of soil salinization is mainly based on medium and low resolution satellite images, and the traditional pixel-based classification method is used to describe the details of salinization information and the monitoring accuracy is not high. In this paper, the domestic GF-1 images are used, and the top-down multiscale segmentation technology and object-oriented information extraction technology are used to accurately extract and classify the salinization information under field scales and to compare with traditional classification methods . The results show that the overall accuracy of the object-oriented method and the maximum likelihood method are 92.72% and 84.31% respectively, and the Kappa coefficients are 0.90 and 0.78 respectively. The technology can accurately extract the salinity information under the field scale and has certain application value and development potential in the future research of high-precision salinization monitoring of farmland.