研究目的
To develop a method for land cover classification using SAR amplitude and textural features from multi-temporal Sentinel-1 images.
研究成果
The method combining SAR amplitude and textural features, especially using the contrast parameter for segmentation, is effective for land cover classification, particularly for water bodies. Using the LBM-DE for segment extraction improves overall classification accuracy.
研究不足
The study is preliminary and focuses on a smaller test site around Berlin. Future work includes extending the classification to larger areas in Germany and comparing with supervised approaches like SVM or Random forest classification.
1：Experimental Design and Method Selection:
The study uses a rule-based decision tree (SAR-LC classifier) optimized to include textual information for processing multi-temporal SAR images.
2：Sample Selection and Data Sources:
Sentinel-1 images from ascending and descending orbits covering the area around Berlin, Germany, were used.
3：List of Experimental Equipment and Materials:
Sentinel-1 satellites, SNAP toolbox, Python programming language, GDAL, eCognition software.
4：Experimental Procedures and Operational Workflow:
Images were post-processed (orbital errors correction, radiometric calibration, thermal noise removal, speckle filtering), texture parameters were extracted, and segmentation and classification were performed.
5：Data Analysis Methods:
The effectiveness of texture parameters for land cover classification was analyzed, and classification accuracies were evaluated.
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
