水稻是人类的主要粮食作物之一，及时准确的获取水稻面积分布和时空变化对粮食政策制定具有重要的参考意义。本文围绕“水稻遥感制图”研究主题，首先回顾调研国内外文献资料，系统梳理了水稻的生理生长过程和主要的种植模式。水稻种植在全球主要集中在东南亚地区；从全国范围看，单季稻产区主要位于东北地区和长江中下游地区；双季稻和三季稻产区位于湖南、江西、广东等华南省份。其次，从遥感数据源出发，归纳了目前水稻遥感制图中常用的光学和雷达遥感数据；在重点分析水稻的“光谱-空间-时间”特征的基础上，探讨了水稻遥感制图中典型植被指数；并从传统机器学习和深度学习两个方面总结了现阶段水稻遥感制图的主流方法。然后，从典型机器学习模型、多源遥感数据融合以及遥感计算云平台归纳了水稻遥感制图的应用现状。总结发现目前水稻制图研究存在以下难点：（1）由于数据缺失或相似生长周期植物的存在导致水稻漏分、错分；（2）依赖机器学习模型参数；（3）地形破碎区域或多季、轮作水稻种植地区的制图困难较大；（4）不同地区种植结构不同导致各方法无法在各地区兼容使用。基于此，从水稻物候特征发掘、水稻时序观测数据获取手段、水稻遥感制图空间分辨率改进等方面探讨了水稻遥感制图的发展方向：（1）水稻物候期遥感信号特征从单一化到多样化；（2）覆盖水稻完整生长期的光学时序数据获取；（3）水稻遥感制图空间分辨率提升；（4）多源遥感数据的有效融合和应用。

Rice is one of the main staple food of human beings, timely and accurate access to distribution of paddy rice cropped area and its spatial-temporal variations are of great significance for food policy formulation. Focusing on the research topic of "paddy rice remote sensing mapping", firstly, we systematically summarized the physiological growing process and main cropping patterns of paddy rice based on the review of domestic and foreign literature. Rice cultivation is mainly concentrated in Southeast Asia globally. In China, the single cropping rice production areas are mainly located in the northeastern region and the middle and lower reaches of the Yangtze River. The double cropping and three-cropping rice production areas are located in southern provinces, such as Hunan, Jiangxi, and Guangdong. Then, from the view of remote sensing data sources, the optical remote sensing data and radar remote sensing data commonly used at present in mapping paddy rice with were summarized. Based on the emphasized analysis of ‘spectrum-space-time" characteristics of paddy rice, typical vegetation index in paddy rice mapping was discussed and the mainstream methods of paddy rice mapping were concluded from traditional machine learning and deep learning. Afterwards, the application status of paddy rice remote sensing mapping was summarized from three aspects: the typical machine learning model, multi-source remote sensing data fusion and remote sensing computing cloud platform. It is concluded that the following issues exist in the current research on rice mapping: (1) Rice is misclassified due to the lack of remotely-sensed data or the similar plants with comparable growth cycle; (2) It is dependent on machine learning model parameters; (3) Rice mapping is difficult in broken terrain or paddy fields subject to multiple seasons or rotation; (4) Different regions have different planting structures, which make it difficult for existing methods to be compatible and applicable universally. Based on this, the development direction of paddy rice mapping was explored from the aspects of rice phenological feature mining, acquisition methods of time-series observations on paddy rice, and improvement of finer spatial resolution of paddy rice mapping: (1) the characteristics of remote sensing signals in rice phenological period have changed from single to diversified; (2) time series optical imagery acquisition covering the entire growth period of rice; (3) improving the spatial resolution of paddy rice mapping; (4) effective fusion and application of multi-source remote sensing data.