Abstract: Accurate acquisition of important information such as planting area and spatial distribution of ginger is essential for market price prediction, yield estimation, disaster warning and policy formulation. However, there is still a lack of research on remote sensing identification of ginger in the Laiwu area under complex cropping patterns. In this study, the planting characteristics and cultivation management characteristics of ginger were analyzed in depth and two ginger indices were constructed to participate in the importance ranking of feature factors, based on Sentinel-2 data, Random Forest classification algorithm and feature preference model were used to extract ginger planting area and spatial distribution. The results indicate that the application of Sentinel-2 data and the constructed ginger index has achieved good results in the remote sensing monitoring of ginger planting area in the Laiwu District, with an overall classification accuracy of 91.32%. Compared with the actual annual planting area, the error is 4.47%, which can meet the practical needs of ginger production and operation. This study has significant reference value for the extraction of ginger planting areas with complex field layouts and diverse cultivation methods.

Key words: ginger area, Sentinel-2, ginger index, feature preference, Random Forest

摘要： 准确获取生姜的种植面积和空间分布等重要信息对于预测市场价格、估算产量、灾害预警和政策制定至关重要。然而，针对复杂种植模式下莱芜地区的生姜遥感识别研究仍然相当匮乏。对生姜的种植特征及栽培管理特点进行深入分析，并构建两个生姜指数用以参与特征因子重要性排序；基于Sentinel-2数据，采用随机森林分类算法和特征优选模型提取生姜种植面积和空间分布。结果表明：Sentinel-2数据和构建的生姜指数的应用对莱芜区生姜种植面积遥感监测有良好效果，总体分类精度可达91.32%，与实际年种植面积相比，误差为4.47%，可满足生姜生产经营的实际需要。对提取具有复杂田间布局和多样化种植方式的生姜种植面积具有重要的参考价值。

关键词: 生姜面积, 哨兵二号, 生姜指数, 特征优选, 随机森林