AUTHOR=Chen Zhaoqi , Wang Chuansheng , Zhang Fuquan , Zhang Ling , Grau Antoni , Guerra Edmundo 

TITLE=All-in-one aerial image enhancement network for forest scenes

JOURNAL=Frontiers in Plant Science

VOLUME=Volume 14 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2023.1154176

DOI=10.3389/fpls.2023.1154176

ISSN=1664-462X

ABSTRACT=Drone monitoring plays an irreplaceable and vital role in forest firefighting because of its wide-range observation angle and real-time transmission of fire characteristics. However, aerial images are often prone to different degradation problems before visual detection. Most current image enhancement methods aim to restore images with specific degradation types. But these methods lack certain generalizability to different degradations, and they are challenging to meet the actual application requirements. Thus, a single model is urgently needed in forest fire monitoring to address the aerial image degradation caused by various common factors such as adverse weather or drone vibration. This paper is dedicated to building an all-in-one framework for various aerial image degradation problems. To this end, we propose an All-in-one Image Enhancement Network(AIENet), which could recover various degraded images in one network. Specifically, we design a new multi-scale receptive field image enhancement block, which can better reconstruct high-resolution details of target regions of different sizes. In particular, this plug-and-play network design enables it to be embedded in any deep learning model. And it has better flexibility and generalization in practical applications. Taking three challenging image enhancement tasks encountered in drone monitoring as examples, we conduct task-specific and all-in-one image enhancement experiments on a synthetic forest wildfire smoke dataset. The results show that the proposed AIENet outperforms or approaches SOTAs quantitatively and qualitatively. Furthermore, in order to prove the effectiveness of the model in advanced vision tasks, we further apply its results to the forest fire detection task, and the detection accuracy has also been significantly improved.