An Automatic Decision-Level Fusion Rice Mapping Method of Optical and SAR Images Based on Cloud Coverage

Abstract

Timely and accurate mapping of paddy rice cultivation is crucial for estimating rice production and optimizing land utilization. Optical images are essential data source for paddy rice mapping, but it is susceptible to cloud contamination. Existing methods struggle to effectively utilize clear-sky pixel information in optical images containing clouds, which impacts the accuracy of paddy rice mapping under cloudy conditions. To address the abovementioned problems, we propose an automatic decision-level fusion rice mapping method of optical and synthetic aperture radar (SAR) images based on cloud coverage (the Auto-OSDF method). The method effectively utilizes clear-sky pixels in images containing clouds and leverages the advantages of SAR features in heavily clouded regions. We tested and validated the Auto-OSDF method in Xiangyin County, Hunan Province, and analyzed the impact of different cloud coverage levels (10%–50%) on the accuracy of rice mapping based on this method. The results indicate that, as cloud coverage increases, the rice mapping accuracy of the Auto-OSDF method is not significantly affected, with overall accuracy and Kappa coefficients both above 93% and 0.90, respectively. To show the value of the proposed method in large-scale applications, we further mapped paddy rice in the entire Hunan Province, and the overall accuracy and Kappa coefficient were 92.47% and 0.87, respectively. The results obtained by the Auto-OSDF method show an average R² of 0.926 compared to municipal-level statistical planting areas. The abovementioned study demonstrates that the Auto-OSDF method is capable of achieving stable and high-precision rice mapping under cloud contamination interference.