哨兵2号多时相植被指数作物分类及监测

Crop classification and monitoring of sentinel 2 multi temporal vegetation index

针对单实相遥感数据分辨率不高以及“同物异谱”和“异物同谱”导致的地物错分问题,依据多时相NDVI植被指数建立决策树判别规则对农作物进行精细分类及面积提取,并对来年种植趋势进行预测。通过监测样本点3—10月的NDVI植被指数变化情况,确定分类阈值并构建决策树分类方法,计算8幅影像的NDVI植被指数,利用决策树分类方法对9种地貌类型进行分类,并对比农作物两年种植面积变化规律,对来年的种植进行分析和预测。实验结果表明,两年的kappa系数分别为0.885 8和0.910 0,总体分类精度分别为90.76%和92.54%,9种地貌类型的遥感分类总体在精度上达到高度一致。

In view of the low resolution of single real-phase remote sensing data and the misclassification of ground objects caused by “same object different spectrum” and “foreign object same spectrum”,this paper establishes the decision tree discrimination rules based on multi temporal NDVI vegetation index, carries out fine classification and area extraction of crops, and forecasts the planting trend in the coming year.By monitoring the change of NDVI vegetation index from March to October, the classification threshold is determined and the decision tree classification method is constructed.Then the NDVI vegetation index of 8 sentry 2 images was calculated.The decision tree classification method is used to classify 9 geomorphic types, compare the change law of crop planting area in two years, and analyze and predict the planting in the coming year.The experimental results show that the kappa coefficients of two years are 0.885 8 and 0.910 0 respectively, and the overall classification accuracy is 90.76% and 92.54% respectively.The overall accuracy of remote sensing classification of 9 geomorphic types is highly consistent.