基于多时相遥感影像的作物种植信息提取

为了快速、准确地在遥感影像上对作物种植信息进行提取,该研究运用多时相的TM/ETM+遥感影像数据和13幅时间序列的MODISEVI遥感影像数据,采取基于生态分类法的监督分类与决策树分类相结合的人机交互解译方法,建立决策树识别模型,对黑龙港地区的主要作物进行遥感解译,总体分类精度达到了91.3%,与单纯对TM影像进行监督分类相比,棉花、玉米、小麦、蔬菜4类作物的相对误差的绝对值分别降低了1.3%、20.5%、2.0%、13.8%。结果表明该方法的分类精度高,能较好的反映作物的分布状况,可为该地区主要作物种植结构调整提供科学依据,还可为其他区域尺度作物分布信息的提取提供参考。

Crops planting information extraction based on multi-temporal remote sensing images

The multi-temporal remote sensing data were used to extract crops planting information quickly and accurately from TM/ETM+ remote sensing images and thirteen MODIS time series remote sensing images, together with the supervised classification and decision tree classification system to interpret major crops in the Heilonggang area. Overall, classification accuracy was up to 91.3%. Compared with one simple supervised classification of TM images, the relative errors of cotton, maize, wheat and vegetables reduced by 1.3%, 20.5%, 2.0% and 13.8% respectively. It proved that this method has high accuracy and it is a good index for the crop planting distribution. The data can provide important scientific information for the adjustment of the major crops planting structure in Heilonggang area and application references for crops classification and crop planting extraction in other area.