A deep learning-based automated Solar-Powered Fish Monitoring System

Abstract

Green fish farming represents an integrated aquaculture approach that rears aquatic organisms in controlled environments to improve production efficiency and environmental sustainability. Although significant, current green fish farming practices are labour-intensive and expensive due to grid energy dependency resulting in operational inefficiencies and elevated fish mortality. To address these key challenges, we propose a multidisciplinary approach that involves the development of a cost-effective, solar-powered automation system that integrates computer vision and deep learning techniques for real-time monitoring of fish behaviour, water quality, feeding, and waste management. First, we design the system architecture that enables automation and ensures accurate system performance under varying conditions. Second, following the architecture, we build a complete and cost-effective smart system that works along with an intelligent software framework that leverages computer vision and deep learning techniques. Utilizing custom datasets from video frames and environmental sensors, this system utilizes convolutional neural networks (CNNs) for fish behavior analysis, real-time disease detection via camera feeds, and precise feeding control through actuators. The design also incorporates a renewable energy subsystem, employing advanced photovoltaic panels and efficient battery storage to guarantee reliable power. The major contribution lies in the seamless integration of these multidisciplinary components. Furthermore, the system architecture is modular and scalable, making it suitable for both smallholder and commercial fish farms. Cost optimization with low-cost sensors and open-source software enables economic viability for resource-constrained farmers. Extensive simulation studies confirmed significant improvements in monitoring accuracy, reduced manual intervention, and enhanced operational sustainability.