Computer Vision-Based Autonomous Underwater Vehicle with Robotic Arm for Garbage Detection and Cleaning

Underwater garbage collection poses a critical issue today, affecting the entire biological chain. Aquatic animals consume suspended plastic, leading to their deaths and disrupting the ecosystem. Approximately eight metric tons of plastic enter the oceans annually, contaminating drinking water with harmful microplastics. Cleaning up this waste remains a major challenge. This paper presents an autonomous underwater vehicle (AUV) designed to address this problem. Equipped with cameras, robotic arms, and a compressor unit, the AUV captures and processes underwater scenes using a deep learning model for garbage detection. Detected garbage is collected by the robotic arms and compressed for increased capacity. The proposed system employs stereovision and triangulation for distance calculation, while the software utilizes transfer learning on a customized pre-trained model, achieving 93% accuracy in object detection and classification of five garbage categories. The system incorporates a dataset of 1200 annotated images, which is further augmented to 1:8 proportion for model development. Overall, this integrated architecture enables efficient underwater garbage detection, collection, and compression.