基于多时相TM和北京一号卫星影像的春播进度遥感监测

基于作物生长的物候规律,利用2007年4月26日、2007年5月28日Landsat TM影像和2007年6月28日北京一号卫星影像进行北京地区春播进度遥感监测。首先,分析了地物类型之间的光谱可分性距离;其次,采用逐步鉴别分析方法,并将掩膜技术和决策树分类方法相结合,监测北京2007年5月28日和6月28日的春播作物种植面积;最后,基于地面调查点对分类结果进行精度评价。结果表明5月28日总体精度为84.5%,6月28日总体精度为88.0%;逐步鉴别分析方法有利于寻找作物分类的光谱差异,建立多时相分类规则,简化了多时相多作物遥感分类流程并提高了分类精度。

Monitoring the Spring Sowing Progress in Beijing with Multi-temporal Landsat TM and "Beijing-1"Micro-satellite Images

The normalized distance between the means was firstly calculated,the phenological spectral differences of multi-crop were investigated by the stepwise discriminations analysis method,the decision tree method and the mask technology were employed to monitor the planting areas of spring-sown crops on 28th May and 28th June,using remote sensing images acquired on 26th April,28th May and 28th June 2007,then the areas in counties were computed with the administrate boundaries,last,the accuracy validations were made based on ground truth interpreting points.The results showed that the classification accuracy on 28th May is 84.5% and that on 28th June is 88.0%.The stepwise discriminations analysis was testified to be more suitable to built classification rule and improve the multi\|crop classification accuracy based on multi\|temporal images.