论文部分内容阅读
利用高光谱遥感技术识别农作物类型已经成为高光谱遥感研究的热点领域。以青海省湟水流域内油菜、小麦和青稞等典型农作物为分类对象,以HJ-1A HSI高光谱数据和GF-1 WFV高分辨率数据为数据源,探讨利用高光谱遥感影像进行农作物类型信息提取的方法。数据经预处理后,首先,利用WFV数据采用面向对象方法提取研究区农作物种植边界,并利用其对HSI高光谱影像进行种植区域提取;其次,将提取后的高光谱影像经数据形式变换获得包括:R、1/R、Log(R)、d(R)、d(Log(R))和CR共6种数据形式;最后,利用上述6种数据形式的全波段数据和经遗传算法GA-SVM进行光谱波段选取后的6种特征数据,采用支持向量机SVM方法进行农作物分类。结果表明:采用基于样本的面向对象分类方法提取耕地信息精度高且实现周期短;利用GA-SVM波段选取后的6种特征数据集进行农作物分类,其精度显著高于全波段数据集分类精度;6种数据变换形式中,d(Log(R))和CR是两种最优的高光谱分类数据形式,其全波段和特征波段数据进行农作物分类均能获得较好的分类精度,总体精度最高分别达88%和86%,而采用1/R、Log(R)和R数据形式需经GA-SVM光谱波段选取后才能获得较优分类精度。
Identifying crop types using hyperspectral remote sensing technology has become a hot area in hyperspectral remote sensing research. Taking HJ-1A HSI hyperspectral data and GF-1 WFV high-resolution data as data sources, the typical crop of rapeseed, wheat and barley in Huangshui River Valley of Qinghai Province was used as data source to investigate the extraction of crop type information using hyperspectral remote sensing images Methods. After preprocessing the data, the WFV data was used to extract the crop planting boundary in the study area by using the object-oriented method, and the HSI hyperspectral image was used to extract the planting area. Secondly, the extracted hyperspectral image was transformed by data transformation to obtain (R), log (R), d (Log (R)), and CR. Finally, using the data of all six bands and the genetic algorithm GA- SVM spectral bands after the selection of six kinds of characteristic data, the use of support vector machine SVM method for crop classification. The results showed that the accuracy of cropland information extraction based on sample-based object-oriented classification was short and the implementation period was short. Using the six characteristic data sets selected by GA-SVM band for crop classification, the accuracy was significantly higher than that of full-band data set classification. Of the six data transformation forms, d (Log (R)) and CR are the two best forms of hyperspectral classification data. The classification accuracy of the whole-band and feature-band data for crop classification can achieve better classification accuracy with the highest overall accuracy 88% and 86% respectively, while the 1 / R, Log (R) and R data formats need to be selected by the GA-SVM spectral band to obtain better classification accuracy.