研究目的
Investigating the effectiveness of deep semantic hashing (DSH) for remote sensing image retrieval by mining semantic information to improve retrieval accuracy and speed.
研究成果
The DSH method significantly improves the detection accuracy of image retrieval by incorporating both visual and semantic features, outperforming other baselines especially in large archives. Future work includes applying the method to other RS images and further optimizing the algorithm by mining semantic information.
研究不足
The archives used are benchmarks with a moderate number of images, which may not fully represent the challenges of much larger archives in real applications.
1：Experimental Design and Method Selection:
The study employs a deep learning-based approach combining feature learning and hash-codes learning in an end-to-end framework.
2：Sample Selection and Data Sources:
Two datasets are used: one from GF-2 satellite and Google Earth, and the CIFAR-10 dataset.
3：List of Experimental Equipment and Materials:
NVIDIA GTX1080 GPU server and MatconvNet for implementation.
4：Experimental Procedures and Operational Workflow:
The method involves feature learning using CNNs for images and semantic annotations, followed by hashing learning to generate binary codes.
5：Data Analysis Methods:
Mean average precision (MAP) is used to measure accuracy.
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
