Surveillance System for Illegal Fishing Prevention on UAV Imagery Using Computer Vision

Abstract

Indonesia has a vast ocean with an abundance of fishes with its natural environments. Those abundances have to be conserved to prevent further destruction of the environment, which can result in the extinction of the surrounding living things. The government had deployed a vessel monitoring system, but illegal fishing still hardly been controlled. In this paper, toward conserving the fishes and especially the environment, we present a surveillance system framework from aerial images using drones technology. We develop a surveillance system using only visual information from the camera installed on the UAV and the design of the convolutional layer for accurate detection. Parameters are learned automatically without manually hand-tuned parameters because the learning process is pure from visual data that learned, so that makes the surveillance and investigation process easier. Experiment show relatively well that the proposed method successfully reaches Average Precision (AP)=75.03%, and hull plate classification reaches Average Matching Precision (AMP)=96.44%, and we believe it could bring many benefits for the ministry of fisheries and marine affairs Indonesia for identifying the illegal vessels and reduce the number of illegal fishing.