基于改进指数特征的渭南市冬小麦面向对象提取

Object Oriented Extraction of Winter Wheat in Weinan City Based on Improved Index Features

准确获取冬小麦种植面积对当地经济发展及规划具有重要意义。本研究采用多时期Sentinel-2影像作为数据源进行时序分析,监测渭南市冬小麦生长变化情况。本研究以Sentinel-2影像3个红边波段分别计算归一化红边指数NDVI_(re 1)、NDVI_(re 2)、NDVI_(re 3),并与其他指数进行面向对象的多特征构建,然后带入到面向对象分类器中进行多特征分类提取,以此获取渭南市精确的冬小麦种植面积。结果表明:基于Sentinel-2影像的植被时序变化分析在渭南市冬小麦生长变化中具有良好的适用性及监测效果,较显著地表明了农作物生长变化情况,并且通过波段组合计算得出的3个红边指数与植被指数均具有良好的相关性,且扩充特征空间后的面向对象分类整体精度为98%,相较于传统的随机森林分类方法有较大的精度提升。

It is of great significance for local economic development and planning to accurately obtain the planting area of winter wheat.In this study,multi-period sentinel-2 images were used as data source for time series analysis to monitor the growth and change of Winter Wheat in Weinan City.In this study,the normalized red edge index NDVI_(re 1),NDVI_(re 2)and NDVI_(re 3)are calculated from three red edge bands of sentinel-2 images,and the object-oriented multi feature construction is carried out with other indexes,and then the multi feature classification is carried out in the object-oriented classifier,so as to obtain the accurate winter wheat planting area in Weinan City.The results show that:the analysis of vegetation time series change based on sentinel-2 image has good applicability and monitoring effect in the growth and change of Winter Wheat in Weinan City,which shows the crop growth and change significantly.The three red edge indexes calculated by band combination have good correlation with vegetation index.And the overall accuracy of object-oriented classification is 98%after expanding the feature space.Compared with the traditional random forest classification method,the accuracy is greatly improved.