Extracting planting area of rapeseed in Sichuan Province based on phenological characteristics of time-series remote sensing images
Crop classification based on high-resolution satellite imagery has important applicative value for agricultural estimation.However,satellite imaging is not effective in areas with frequent cloud cover and rain,and single-phase optical imagery cannot be used to accurately analyze the patterns of growth and differences in the features of different crops.In addition,the spectral similarity of the vegetation is high,and changes dynamically over time and space,where this poses a major challenge to high-precision crop identification.This study uses the exclusivity and stability of the phenological features of crops as the starting point,and uses an optimized three-harmonic model of fitting to propose a framework for the multi-feature classification of crops.It integrates such auxiliary information as spectral,topographic,and textural data.The authors used this framework along with the Google Earth Engine to identify and extract the planting area of rapeseed in Sichuan Province in 2020 and 2021 by using remote sensing images.The results showed that the proposed framework was able to extract the planting area of rapeseed with an overall classification accuracy of 96.6％and a Kappa coefficient of 0.906.Its results of classification were in good agreement with data from statistical yearbooks.It has a higher accuracy of classification than prevalent methods in the area,and can quickly extract the fine spatial distribution of crops over multiple phases and large areas.
harmonic analysisGoogle Earth Enginephenological featurestime seriesspatial distribution
万方数据
维普
