Work

Classification of paddy field water conditions with unmanned aerial vehicle: A novel method incorporating the short-wave infrared band

Year: 2026

Type: article

Abstract: Monitoring paddy field water conditions (PFWC) presents unique challenges due to frequent alternate wetting and drying cycles and complex canopy conditions. Existing studies on unmanned aerial vehicle (UAV)-based field water condition identification have integrated multi-source information, including thermal infrared (TIR), which limits the monitoring time to a relatively short period and reduces data acquisition and processing efficiency. Furthermore, their models have rarely been evaluated in complex field conditions. This study utilized UAV data acquired from a multispectral (MS) including short-wave infrared (SWIR) + thermal infrared (TIR) sensor to develop several PFWC classification models, then selected the optimal model and information source, and verified the model’s generalizability across multiple scenarios. The results indicated that: (1) MS including SWIR source achieved satisfactory PFWC classification performance over the 9:00–15:00 time window, with average Accuracy= 0.918, Macro-F1= 0.899 for the optimal model (Support Vector Machine, SVM) across three growth stages, whereas MS + TIR excluding SWIR was only effective over the 12:00–15:00 time window. Meanwhile, adding TIR to MS including SWIR cannot effectively improve classification accuracy. (2) Multi-scenario testing showed, except for suboptimal performance in paddy lodging scenario, our PFWC classification method performed well across fields under practical farming conditions by farmers, insufficient nitrogen application, and pest infestation, demonstrating strong cross-scenario generalization. Overall, using MS including SWIR as input to SVM is an effective UAV-based PFWC classification method that extends the operational monitoring window compared with TIR, improves data acquisition and processing efficiency, and provides useful support for the development of intelligent irrigation forecasting in paddy irrigation areas.

Source: Agricultural Water Management

DOI: 10.1016/j.agwat.2026.110626