研究目的
Investigating the efficiency and accuracy of a multilevel change detection method combining unsupervised object-based correlation analysis and supervised post-classification comparison in optical remote sensing images.
研究成果
The proposed approach outperforms state-of-the-art methods in terms of accuracy and automation degree, demonstrating its effectiveness and robustness in change detection for optical remote sensing images.
研究不足
The method requires the selection of training examples for supervised algorithms, and the segmentation parameter w needs to be chosen carefully to avoid over-segmentation.
1：Experimental Design and Method Selection:
The methodology involves fast multitemporal segmentation for object-level information, followed by object-based correlation analysis (OBCA) for extracting potential changed areas, and post-classification comparison (PCC) for refining results.
2：Sample Selection and Data Sources:
Multitemporal images taken by the Chinese Gaofen-2 satellite with a size of 500×500 pixels and spatial resolution of 1m/pixel are used.
3：List of Experimental Equipment and Materials:
The watershed method for segmentation, SVM for classification, and a combination of GLCM, HOG, mean value, and standard deviation as descriptors.
4：Experimental Procedures and Operational Workflow:
Segmentation is applied on two registered images, followed by OBCA to extract changed areas, and PCC to refine the change map.
5：Data Analysis Methods:
Performance is evaluated using false alarms (FA), missed alarms (MA), overall alarms (OA), and kappa coefficient.
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
