Feature Enhancement Method for Text Information Generation of Remote Sensing Images
Qihui Lin, Sen Li, Lurui Xia, Yihe Shan · 2022 4th International Conference on Intelligent Control, Measurement and Signal Processing (ICMSP) · 2022
The redundancy of irrelevant information brings great inconvenience to the interpretation of remote sensing images. Image captioning can efficiently extract effective information and eliminate irrelevant data to the greatest extent. However, inadequate use of visual features is a common problem in the current research on remote sensing image captioning. To solve this problem, we propose two visual feature enhancement methods for remote sensing image captioning, which are the combination of multi-scale features and the use of remote sensing images pre-training weights. Experiments are carried out using different language models, and the results prove the effectiveness of our proposed feature enhancement methods.