论文部分内容阅读
在区域或全球尺度,250m分辨率的MODIS EVI常被用于作物分类。而且,基于遥感数据可以快速准确的进行作物分类,并为辅助农业政策的制定,因而得到了广大研究者的关注。研究提出了直接使用多年MODIS 250mEVI和临近年份地面调查数据进行作物分类的方法。首先,通过扩展2011,2012和2013年的野外调查数据获得全疆的典型地块,并从地块中提取MODIS纯像元作为分类样本。接着使用免疫系统网络分类器(ABNet)提取研究取的主要作物,包括棉花、玉米、冬小麦和葡萄等。在三年的数据中,任意两年的地面数据用于训练分类器,用使用训练好的分类器对另一年的数据进行分类。例如,使用2011和2012年的数据训练分类器,并对2013年的数据进行分类。结果表明,每年的分类精度达80%以上,且Kappa系数高于0.7。今后工作中,仍需使用更多的地面数据获得更的更精细的分类结果。
MODIS EVIs of 250m resolution are often used for crop classification at the regional or global scale. Moreover, based on remote sensing data, crop classification can be carried out quickly and accurately, and to assist in the formulation of agricultural policy, it has attracted the attention of researchers. The study proposed a method of crop classification using the MODIS 250mEVI directly for years and the ground survey data in recent years. First of all, by expanding the field survey data of 2011, 2012 and 2013 to obtain the typical plots in Xinjiang, MODIS pure pixels were extracted from the plots as the classified samples. Then use the immune system network classifier (ABNet) to extract the study of the main crops, including cotton, corn, winter wheat and grapes. In three years of data, any two years of terrestrial data are used to train the classifier, and the trained classifier is used to classify another year of data. For example, classifiers are trained using data from 2011 and 2012, and data for 2013 are categorized. The results show that the classification accuracy is more than 80%, and the Kappa coefficient is higher than 0.7. In future work, more ground data still need to be used to obtain more refined classification results.