Computer Vision Corrections Enhance UAV-Based Retrievals in Shallow Waters

Demo

Traditional methods for shallow water (<5 m) mapping are expensive and spatially constrained due to the complexity of sensor deployment in remote coastal zones. These limitations hinder wide-scale monitoring of ecologically critical near-shore benthic habitats.

This project leverages Unmanned Aerial Vehicles (UAVs) to collect high-resolution video data and introduces a novel processing pipeline to remove optical distortions such as:

• 💡 Light refraction
• 🌊 Caustics
• ✨ Sun glint

In our study, we process Full-HD 60 FPS UAV videos captured at altitudes ranging from 10 m to 120 m. Our approach includes:

• 🎞️ Frame Extraction
• 🧮 Image Averaging (Median Filtering)
• 🎨 Color Transferring Correction

In the final enhancement, we were able to achieve not only the restoration of the geometries, but also a color enhancement to better discriminate the underwater scenary.

Clone the repository and install required packages:

git clone https://github.com/Dark3850/CV_Caustics_Removal.git
pip install -r requirements.txt 

The Python version used in our work is python==3.9.1

1. To download the sample video, use this Drive link.
2. Place the video file in the Caustics_Removal folder.
3. Then simply run the CV_Caustics_Removal.jpynb notebook.

On a general note, you will find the ground truth image to perform the final enhancement correction in the Caustics_Removal/5_Target folder. This image was the one used in our study.

1. Be sure to put all your UAV videos in the Caustics_Removal folder.
2. Put your ground truth image in the Caustics_Removal/5_Target fodler. For more details on the GT iamge selection, look at our paper.
3. Run the CV_Caustics_Removal.jpynb notebook. The folder structure is already given in the repository.

If you use this repository or the methods described herein in your work, please cite it as follows. Link to the article. BibTeX

@ARTICLE{11075526,
 author={Scilla, Dario and Lopez, Omar A. and Nieuwenhuis, Brian Owain and Johansen, Kasper and Elías-Lara, Mariana and Angulo, Victor and Rodríguez, Jorge L. and Jones, Burton H. and McCabe, Matthew F.},
 journal={IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing}, 
 title={Computer Vision Corrections Enhance UAV-Based Retrievals in Shallow Waters}, 
 year={2025},
 volume={18},
 number={},
 pages={18134-18149},
 keywords={Image color analysis;Autonomous aerial vehicles;Optical distortion;Sun;Marine vegetation;Distortion;Videos;Shape;Water conservation;Sea measurements;Caustics;color transferring;computer vision;refraction;shallow waters;unmanned aerial vehicle (UAV)},
 doi={10.1109/JSTARS.2025.3587478}}