Classification of KJA Net Conditions Using ROV and Computer Vision

Abstract

The development and integration of Remotely operated vehicle (ROV) with computer vision has been carried out and shows excellent performance. All ROV features functions run smoothly and without problems and are able to monitor the condition of nets in floating net cages (KJA) and produce underwater videos. Data collected from ROV are processed, utilizing the YOLOv8 model and showed very positive results in classifying the condition of KJA nets. The model achieves an accuracy level of 1 or 100% differentiate between clean and dirty net. Based on these results, it can be concluded that the YOLOv8 model has excellent performance in recognizing mesh objects with a high level of accuracy. These results provide confidence that this model can be trusted in monitoring the condition of KJA nets.