论文部分内容阅读
本文选取土壤质地这一显著影响作物生长的土壤特性,进行卫片微机数字制图试验。试验区包含海宁县大部和部分邻县面积,共计936平方公里.于4个谱段的放大透明卫片上,以疏格网(grid),大量孔采样网格(cell)平均光学密度,输入TRS—80微型机。经特征提取,相继通过自行设计的监督分类器和非监督分类器,实现统计模式识别,并以键盘字符宽行打印图输出分级成果。其中,监督分类平均符合率达75.1%,不亚于以专用遥感图象处理系统取得的单一土壤特性数字分级精度。
In this paper, soil texture is selected as the soil characteristics that significantly affect the growth of crops, and the satellite digital mapping experiment is carried out. The experimental area consists of a large part of Haining County and part of neighboring counties with a total area of 936 square kilometers.On magnified transparent panels of 4 spectral bands, the average optical density of the grid, large number of hole sampling cells, TRS-80 microcomputer. After feature extraction, the self-designed supervised classifier and unsupervised classifier are used successively to realize the statistical pattern recognition, and the graded results are output with the wide-character print of keyboard characters. Among them, the average coincidence rate of supervised classification reaches 75.1%, which is no less than the digital grading precision of single soil characteristic obtained by special remote sensing image processing system.