FY－3C／MERSI 卫星影像水稻信息自动提取的决策树方法研究

摘要

为了推进风云三号系列卫星资料在水稻信息提取中的应用，分析了FY－3C／MERSI影像中江西省区域水稻及其他典型地物的光谱特征及通道间的关系，并讨论了各地物在光谱特征上的可分性。研究发现：除植被外，其他地物均可依据适当的阈值与水稻分开；水稻与其他植被在各单通道上的光谱特性相似，因此引入归一化植被指数来提高水稻提取的精度。根据以上研究结果，建立了决策树模型对水稻信息进行自动提取。目视检验结果表明，该方法的提取效果较好，只是在水稻与其他植被交界处有误判现象。提取精度评价结果显示，该方法的总体提取精度较高，基本上可以满足水稻生育期面积遥感监测与产量预报的需求。%In order to promote the application of FY-3 series of satellite data in information extraction of paddy rice, the au-thor analyzed the spectral characteristics of paddy rice and other typical objects and the relationship among channels on FY-3C/MERSI image in Jiangxi province, and discussed the separability around objects on the spectral characteristics.The results showed the other typical objects could be separated from paddy rice on the basis of appropriate threshold besides vegetation.The spectral characteristics of paddy rice on every single channel was similar with the other vegetation, so the Normalized Difference Vegetation Index of vegetation was introduced to improve the extraction accuracy of paddy rice.According to the results of the research, the decision-tree model was established to automatically extract paddy rice information.Visual inspection showed that the extracting effect of this method was better, however, some pixels in the neighborhood area between paddy rice and other vegetation were judged by mistakes.Precision evaluation results showed that the overall extraction precision of this method was higher, the extrac-tion results could meet the demand of remote sensing monitoring of rice area in the growth period and yiejd forecast.