极化SAR 农作物分类研究进展

Research advances on crop classification using PolSAR data

关键词:  全极化合成孔径雷达  农作物分类  极化目标分解

Key words:  Polarimetric SAR（PolSAR）  Crop classification  Polarimetric target decompositions

摘要:

【目的】农作物分类是农情遥感监测的重要环节。及时、准确地获取农作物类型、面 积及空间分布信息对加强农业生产管理、制定经济政策以及保障国家粮食安全具有重要意 义。【方法】文章从监测的农作物类型、使用的数据源、分类特征、算法及精度等方面系统 总结了近10 余年来农作物极化SAR 分类的研究进展，梳理归纳了农作物SAR 分类特征及 其提取方法，对比分析了各种极化SAR 分类方法的优缺点及适用条件，并总结了多源多时 相数据在极化SAR 农作物分类中的应用。【结果】利用极化SAR 数据进行作物分类的精度 逐步提高，但仍存在以下不足：早期极化SAR 监测的作物类型较为单一，以水稻为主，近 期虽涉及多种作物类型，但针对个别旱地作物的分类研究精度仍不高；针对农作物尤其是旱 地作物的散射机制研究明显不足，适合各种旱地作物的分类特征尚未明确与优选；农作物极 化SAR 分类算法以统计方法和机器学习算法为主，机理性和稳定性不强。【结论】农作物极 化SAR 分类未来的发展方向：（1）深入研究农作物散射机制，发展更具普适性的分类算法； （2）选取用于分类的关键时相、关键特征；（3）多源数据融合，充分发挥各自优势，提高分 类精度。

Abstract:

［Purpose］Crop classification is an important part of agricultural remote sensing monitoring. Timely and accurate access to crop type，area and spatial distribution information is of great significance for strengthening agricultural production management，formulating economic policies and ensuring national food security.［ Method］The research progress of crop polarimetric SAR（PolSAR）classification in recent 10 years is systematically summarizes from the aspects of crop types，data sources，classification features，algorithms and accuracy. The classification features and extraction methods of crop PolSAR classification are summarized. The advantages and disadvantages of various PolSAR classification methods and their applicable conditions are compared and analyzed. The application of crop PolSAR classification using multisource and multi-temporal data is summarized. On this basis，the shortcomings of current crop PolSAR classification are pointed out，and the future research directions is prospected. ［Result］Although the accuracy of crop classification using polarimetric SAR data has been gradually improved，the following deficiencies still exist：first，the crop types monitored by polarimetric SAR in the early stage are relatively single，mainly rice. Many crop types are involved in the near future，but the classification accuracy of dryland crops is still not high. Secondly， the backscattering mechanism of crops，especially dryland crops is obviously insufficient，and the classification characteristics suitable for various dryland crops have not been defined and optimized. Finally，most of the PolSAR crop classification algorithms are statistical methods and machine learning algorithms，which have poor mechanism and stability.［ Conclusion］The future research directions of crop polarimetric SAR classification are：（1）In-depth research on the scattering mechanism of crops to develop more universal classification algorithms. （2）Selection of key phases and key features for PolSAR crop classification.（ 3）Fusion of multisource data to give full play to their respective advantages and improve classification accuracy.