基于Sentinel-1/2数据一种改进的决策树方法提取荔枝果园研究

An Improved Decision Tree Method for Extracting Litchi Orchards Based on Sentinel-1/2 Data

荔枝种植在热带、亚热带地区,受限于多云多雨天气和遥感技术,获取多时相并且能够表征荔枝物候特征的有效高光谱与高空间分辨率数据是比较困难的,这阻碍了荔枝果园的提取。采用Sentinel-1时间序列微波数据构建了荔枝关键生育期特征,并结合Sentinel-2数据提出了一种改进的荔枝果园识别模型,并将该模型应用于茂名市荔枝果园的识别。结果表明:成熟期荔枝与其他植被或作物的VH极化特征差异显著;该方法在荔枝果园空间分布制图的总体精度为90.34%,Kappa系数为0.84。该方法能够有效刻画荔枝的关键生育期特征,且准确识别光谱与空间特征相似的荔枝果园。为热带和亚热带其他果园的识别提供了参考,对果园的生长监测、精确管理和产量估算具有重要意义。

Litchi cultivation is limited by cloudy and rainy weather as well as remote sensing technology in tropical and subtropical regions,which makes obtaining multi-temporal,and effective hyperspectral and high spatial resolution data to characterize Litchi's phenology difficult.As a result,the extraction of Litchi orchards is hindered.In this research,Sentinel-1 time-series microwave data was utilized for constructing a crucial fertility characteristic of Litchi.An enhanced methodology for Litchi orchard identification was developed by incorporating Sentinel-2 data.The model was subsequently utilized to identify Litchi orchards in Maoming City.The study revealed that the VH polarization properties of mature Litchi distinguish significantly from those of other crops and vegetation.The overall accuracy of spatial distribution mapping in Litchi orchards was 90.34%,and the Kappa coefficient was 0.84.This method can effectively characterize the key growth stage of litchi,and accurately identify Litchi orchards with similar spectral and spatial characteristics.This study serves as a valuable reference for identifying orchards in the tropics and subtropics.Its significance lies in enabling accurate monitoring of growth,managing orchards precisely,and estimating yields.