Implementation of YOLO V5 Integrated on Jetson Nano for Autonomous Underwater Vehicle Prototype to Overcome Pollution in Aquatic Areas

Indonesia is a country with vast maritime territory. According to the Indonesian Central Statistics Agency (BPS) in 2022, Indonesia consists of 17,001 islands spread from Sabang to Merauke. Based on this report, it is evident that Indonesia possesses extensive maritime areas, including seas, straits, bays, and rivers. However, Indonesia faces water pollution issues. To address this, Autonomous Underwater Vehicle (AUV) technology is utilized, supported by image processing algorithms like Yolo V5 to accurately detect and identify water pollution. Leveraging these advancements, robots are designed using YOLO V5 to gather object class datasheets for accurate classification. Similarly, after successfully detecting objects, robots determine movement positions using Proportional-Integral-Derivative (PID) to ascertain center of bounding box accuracy, orientation stability, and completion time. Research results indicate that handling 1300 to 300 object numbers can achieve high average precision with accurate confidence. This serves as a reference for robots in determining movements towards objects using error and desired value comparisons. PID testing on robots shows that with Kp:Ki:Kd settings of 50:5:0.5, Center of Bounding Box Accuracy of 85.4 and Orientation Stability of 79.4 are achieved in 4.8 seconds. With settings of 100:5:2, these values increase to 91.4 and 89.6 with an allocation time of 3.4 seconds. These conditions demonstrate that increasing PID values contributes to improved robot performance, characterized by faster allocation times and higher stability. With settings of 125:12:8, Center of Bounding Box Accuracy of 96.4 and Orientation Stability of 90.8 are achieved in 2 seconds, enhancing robot stability further.