研究目的
To propose a more efficient CNN model for extracting deep features of remote sensing images and construct a graph model based on these features for precise built-up area extraction.
研究成果
The proposed CNN model and graph-based method effectively extract built-up areas from remote sensing images, demonstrating superior performance and speed compared to traditional methods. The approach is balanced in accuracy metrics and computationally efficient.
研究不足
The results are block-level, which may not capture finer details at the pixel level. Future research could address this limitation.
1：Experimental Design and Method Selection:
The study designs a CNN model with few parameters and computation for extracting deep features from remote sensing images. A graph model is then constructed based on these features for built-up area extraction.
2：Sample Selection and Data Sources:
The study uses a dataset of twenty Gaofen-2 satellite images, each including panchromatic and multispectral images, for validation.
3：List of Experimental Equipment and Materials:
The study utilizes deep learning and graph theory techniques without specifying hardware.
4：Experimental Procedures and Operational Workflow:
The process involves dividing the image into blocks, calculating deep features and category probabilities via CNN, constructing a graph model, and applying Max Flow/Min Cut algorithm for segmentation.
5：Data Analysis Methods:
The study evaluates performance using User's Accuracy, Producer's Accuracy, and Overall Accuracy based on a confusion matrix.
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
