AI-Powered Web based Underwater Image Enhancement for Flood Monitoring using CNN GANs
Year: 2025
Type: article
Abstract: Light absorption and dispersion in water create intrinsic distortions that lead to low contrast, colourdeterioration, and blurriness, making underwater picture improvement a crucial task. Underwater imagingis crucial for applications including marine biology, underwater archaeology, and flood monitoring, butthese problems make it more difficult to understand and analyse. Detecting human presence in underwaterenvironments is also essential for safety monitoring and rescue operations. In order to solve this, ourproposal incorporates a person detection model based on YOLO, which allows for precise andinstantaneous identification of people in submerged environments. This feature is especially helpful inflood catastrophe situations, since prompt and accurate detection can help rescue crews find and helpvictims. A Convolutional Neural Network in conjunction with a Generative Adversarial Network (CNN-GAN) is the unique method we suggest to further improve underwater image quality. Images producedthrough this adversarial training process have better colour correction, contrast, and sharpness, whichmakes them aesthetically comparable to photos taken under more favourable lighting conditions. Awebsite that enables individuals to submit low-quality underwater photos and receive improved versionsand human identification is the project's output. Authorities can also use this function to monitorunderwater environments or evaluate areas impacted by flooding. Only authorised individuals can accesssensitive data due to the website's secure login system. The outcomes show notable enhancements in imagequality, which makes the website a useful resource for environmental monitoring andunderwater exploration research as well as real-world applications.
Authors: Archana CV, Leya MS, Shaji B, Justin Jose
Language: English
Topic: Image Enhancement Techniques
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4408886324