Guo Xianyu, Yin Junjun, Li Kun, Yang Jian
School of Computer and Communication Engineering, University of Science and Technology Beijing, Beijing, China.
Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing, China.
Front Plant Sci. 2024 Oct 10;15:1391735. doi: 10.3389/fpls.2024.1391735. eCollection 2024.
Fine classification and phenological information of rice paddy are of great significance for precision agricultural management. General compact polarimetric (CP) synthetic aperture radar (SAR) offers the advantage of providing rich polarimetric information, making it an important means of monitoring rice growth. Therefore, in response to the current challenges of difficulty in rice type classification and the small differences in phenological polarimetric characteristics, a novel strategy for fine classification and phenological analysis of rice paddy is proposed. This strategy thoroughly explores the polarimetric information of general CP SAR data and the target scattering characterization capabilities under different imaging modes. Firstly, the general CP SAR data is formalized using the standard CP descriptors, followed by the extraction of general CP features through the Δ / target decomposition method. optimal CP features are generated to achieve fine classification of rice paddy. Finally, 6 phenological stages of rice are analyzed based on the general CP features. The experiment results of rice classification show that the classification accuracy based on this strategy exceeds 90%, with a Kappa coefficient above 0.88. The highest classification accuracies were observed for transplanting hybrid rice paddy (T-H) and direct-sown japonica rice paddy (D-J), at 80.9% and 89.9%, respectively. The phenological evolution rule of the two rice types indicate that from early June (seedling stage) to late July (elongation stage), the CP feature variation trends of T-H and D-J are generally consistent. However, from October (mature stage) to November (harvest stage), the variation trends of the CP features for T-H and D-J are significantly different. The study found that from the booting-heading stage to the harvest stage, the linear π/4 mode outperforms circular and elliptical polarimetric modes in distinguishing different types of rice. Throughout the entire phenological period of rice growth, CP SAR of linear π/4 mode is surpasses that of other linear modes in discriminating different type of rice. The proposed strategy enables high-precision fine classification rice paddy, and the extracted general CP parameter effectively reflects the phenological change trends in rice growth.
稻田的精细分类和物候信息对于精准农业管理具有重要意义。通用紧凑极化（CP）合成孔径雷达（SAR）具有提供丰富极化信息的优势，使其成为监测水稻生长的重要手段。因此，针对当前水稻类型分类困难以及物候极化特征差异较小的挑战，提出了一种稻田精细分类和物候分析的新策略。该策略深入探索了通用CP SAR数据的极化信息以及不同成像模式下的目标散射表征能力。首先，使用标准CP描述符对通用CP SAR数据进行形式化，然后通过Δ/目标分解方法提取通用CP特征，生成最优CP特征以实现稻田的精细分类。最后，基于通用CP特征分析水稻的6个物候阶段。水稻分类实验结果表明，基于该策略的分类准确率超过90%，Kappa系数高于0.88。移栽杂交稻田（T-H）和直播粳稻田（D-J）的分类准确率最高，分别为80.9%和89.9%。两种水稻类型的物候演变规律表明，从6月初（苗期）到7月底（拔节期），T-H和D-J的CP特征变化趋势总体一致。然而，从10月（成熟期）到11月（收获期），T-H和D-J的CP特征变化趋势明显不同。研究发现，从孕穗抽穗期到收获期，线性π/4模式在区分不同类型水稻方面优于圆极化和椭圆极化模式。在水稻生长的整个物候期，线性π/4模式的CP SAR在区分不同类型水稻方面优于其他线性模式。所提出的策略能够实现稻田的高精度精细分类，提取的通用CP参数有效反映了水稻生长的物候变化趋势。
