Enhancing Marine Debris Acoustic Monitoring by Optical Flow-Based Motion Vector Analysis

With the development of coastal construction, a large amount of human-generated waste, particularly plastic debris, is continuously entering the ocean, posing a severe threat to marine ecosystems. The key to effectively addressing plastic pollution lies in the ability to autonomously monitor such debris. Currently, marine debris monitoring primarily relies on optical sensors, but these methods are limited in their applicability to underwater and seafloor areas due to low-visibility constraints. The acoustic camera, also known as high-resolution forward-looking sonar (FLS), has demonstrated considerable potential in the autonomous monitoring of marine debris, as they are unaffected by water turbidity and dark environments. The appearance of targets in sonar images changes with variations in the imaging viewpoint, while challenges such as low signal-to-noise ratio, weak textures, and imaging distortions in sonar imagery present significant obstacles to debris monitoring based on prior class labels. This paper proposes an optical flow-based method for marine debris monitoring, aiming to fully utilize the time series information captured by the acoustic camera to enhance the performance of marine debris monitoring without relying on prior category labels of the targets. The proposed method was validated through experiments conducted in a circulating water tank, demonstrating its feasibility and robustness. This approach holds promise for providing novel insights into the spatial and temporal distribution of debris.