农作物类型遥感识别算法及国产高分卫星应用示例

关键词:  高分卫星  农作物  遥感识别  算法  应用示例

Key words:  GF satellite  crop  remote sensing recognition  algorithm  application example

摘要:

目的 随着国产高分卫星的快速发展，高分卫星遥感数据在农作物类型遥感识别中获得了广泛应用。通过对农作物类型遥感识别算法的总结，可以发现农作物类型遥感识别算法的优缺点，能够提高国产高分卫星在农业遥感监测中的应用水平。方法 文章在农作物类型遥感识别算法原理分析基础上，对农作物类型遥感识别算法进行分类，并基于高分六号宽幅数据（GF-6/WFV），在各类算法中选择1~2种典型方法进行示例说明。结果 光谱特征的可分性是遥感识别算法的基础，根据实施方案的差异，农作物类型遥感识别算法可分为目视解译算法、机器学习法和特征指数识别算法3类。基于GF-6/WFV的随机森林的花生类作物识别、光谱空间特征指数算法的园地识别和时间序列特征指数的冬小麦识别的结果精度均大于90%。结论 当前农作物类别遥感识别算法中，高分卫星数据都能够得到有效应用，国产高分卫星数据的应用前景广阔。

Abstract:

OF REMOTE SENSING ALGORITHM IN CROP TYPE IDENTIFICATION AND ITS APPLICATION BASED ON GAOFEN SATELLITES^*Ji Fuhua，Liu Jia^※，Wang Limin(Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences, Beijing 100081, China)With the rapid development of domestic GF satellite, GF satellite data has been widely used in the remote sensing identification of crop types. Through the summary of the GF satellite crop type remote sensing recognition algorithm, the advantages and disadvantages of remote sensing recognition algorithms for crop types can be found, which can promote the application level of GF satellite data in agricultural remote sensing monitoring. Based on the principal analysis of the crop type remote sensing recognition algorithm, the remote sensing recognition algorithm of crop type was classified. Also, 1-2 methods from each type of algorithm were further illustrated as example according to the GF-6/WFV data. The results were listed as follows. The separability of spectral features was the basis of remote sensing recognition algorithm, in the consideration of the difference within implementations, remote sensing algorithms in crop types recognition were divided into three categories, including manual visual interpretation and machine learning classification method and feature index algorithm. Using GF-6/WFV data, the accuracy of peanut crop identification based on random forest algorithm, orchard identification based on spectral spatial feature index algorithm and winter wheat identification based on time series feature index algorithm were all greater than 90%, which met the accuracy requirements of crop identification and verified the feasibility of the three types of crop type recognition algorithms based on GF satellite data. To summary, in the current crop type remote sensing recognition algorithms, GF satellite data can be effectively used, and the capabilities of domestic GF satellite data need to be explored in depth.