研究目的
To evaluate the potentiality of polarimetric C- and L-SAR time-series to improve the identification and characterization of vegetation cover during winter season in a 130 km2 area.
研究成果
The study confirms the potential of SAR time series, particularly Radarsat-2 polarimetric images, for identifying land use during winter with an overall accuracy of approximately 80%. However, classification accuracy decreases with increased detail in land-use types, and there are differences between sensors. Future work should explore combining data from multiple SAR sensors to improve precision in monitoring winter land-use, especially for main crops.
研究不足
The study faced limitations in accurately classifying all land-use types, with lower accuracies at more detailed classification levels (e.g., main crops). Differences in accuracy were observed between sensors, with Radarsat-2 performing best but still not perfect. Atmospheric conditions and sensor-specific characteristics (e.g., polarization states, spatial resolutions) may affect results, and the use of combined data from multiple sensors could be optimized for better performance.
1：Experimental Design and Method Selection:
The study used polarimetric SAR time-series from Alos-2, Radarsat-2, and Sentinel-1 satellites to classify land-use during winter. Random Forest (RF) algorithm was employed for classification, with pre-processing steps including radiometric calibration, speckle filtering, geocoding, and extraction of backscattering coefficients and polarimetric parameters such as Shannon Entropy and SPAN.
2：Sample Selection and Data Sources:
Field observations were conducted in 235 crop fields in the Pleine-Fougères LTER site in France from October 2016 to March 2017. Two-thirds of plots were used for training and one-third for validation. Satellite imagery included ten Radarsat-2, eight Alos-2, and nine Sentinel-1 images acquired from September 2016 to June
3：Two-thirds of plots were used for training and one-third for validation. Satellite imagery included ten Radarsat-2, eight Alos-2, and nine Sentinel-1 images acquired from September 2016 to June List of Experimental Equipment and Materials:
2017.
3. List of Experimental Equipment and Materials: SAR images from Alos-2, Radarsat-2, and Sentinel-1; PolSARpro v5.1.1 software for polarimetric processing; SNAP v5.0 software for geocoding; Shuttle Radar Topography Mission (SRTM) data for topographic correction; and ground survey data.
4：1 software for polarimetric processing; SNAP v0 software for geocoding; Shuttle Radar Topography Mission (SRTM) data for topographic correction; and ground survey data.
Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: Images were pre-processed to extract backscattering coefficients and polarimetric parameters, filtered for speckle reduction, geocoded, and classified using RF with K-fold cross-validation for accuracy assessment.
5：Data Analysis Methods:
Classification accuracies were evaluated using overall accuracy metrics derived from cross-validation, comparing results across different sensors and land-use classification levels.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
