So far, many depend on fisheries, particularly in fish farming or aquaculture, as a food source. Previously, fish farming was a small-scale production that only addressed the need for a family or a small community’s livelihood source. The increasing demand for aquaculture industrialization pushed small-scale farmers to improve their farming skills to increase production in their operations. One of the challenges to aquaculture property security is theft and human intrusions that can affect profit and farm operations, particularly to larger aquaculture sites with more financial investments and resources. Additionally, many of the aquaculture site locations are in open water and have higher risks or vulnerability to possible intrusions. Unauthorized vessels could access fish-cage areas for illegal fishing that causes profits losses. Most of the time, discovering a security breach occurs during harvest time. Others rely on on-site security, using barriers and employing guards to monitor and survey, but this entails additional expense from the owners. Aside from security threats, fish welfare and fish feeding behaviors are significant factors for successful fish farming. Fish behaviors have practical and economic significance in fishery production. It has also become an essential theoretical basis to improve or guide fishing and production techniques by applying new technologies .
In the past, fish fed from natural food which came from natural water resources. Due to the development of aquaculture methods, farmers are now adopting artificial fish feeding. Since artificial feeding was adopted, the amount of feeding has become a great concern, so that neither inadequate feeding nor overfeeding occurs. If underfeeding affects the quality and quantity of fish yield, overfeeding decreases the profit of fish farmers due to food waste. Fish overfeeding also deteriorates the environment, particularly in water quality, and can further harm the fishing industry in general. To address this concern, farmers use direct feeding observation and manual human recording to monitor feeding, which is labor-intensive, time-consuming and energy-consuming. Additionally, human observation is prone to subjectivity and errors, and is not suitable as a continuous, accurate, and consistent source of information .
Sub-Saharan African aquaculture sector challenged by feed accessibility
The aquaculture industry is still one of the fastest growing agricultural sectors despite many challenges, and is responsible for providing more available and affordable fish products on the market. The Department of Fisheries and Aquaculture under the Food and Agricultural Organization of the United Nations reported that aquaculture is the leading source of worldwide fish production. Aquaculture’s contribution to the world’s total production remains consistent and rapid, from 25.7% in 2000 to 46.8% in 2016
. This significant progress can be attributed to the farmer’s utilization of aquaculture management software to reduce costs and optimize production . Many technological innovations have been proposed and utilized to improve aquaculture production and management. However, many existing aquaculture management systems have problems with cost, mobility, efficiency, and functionality. Many of these current systems use expensive sensors for data collection. Additionally, site-wide monitoring of the entire aquaculture area requires the installation of multiple fixed cameras. Others use high-priced drones equipped with aerial cameras, sensors, and computational capability on board for site surveillance. Drones are reliable in monitoring, but they have a limited power source. Embedding the computing strength in the drone affects its real-time operations and limits its efficiency, mobility, and decreases navigation time .
Full system features are likewise expensive to acquire and require higher technical skills. The combination of radars, identification systems, drones, hovercrafts, thermal cameras, night vision, sonar, echo sounders, and countermeasures is costly. Our aquaculture surveillance provides physical site monitoring for aquaculture farms in order to monitor unauthorized individuals or ships for possible theft. The surveillance of fish behavior and growth will allow aquaculture farmers to monitor fish growth. Both surveillance methods can help ensure that no profit is lost due to security breaches, and that aquaculture production is optimized using fish feeding intensity, fish count, and fish-length estimation.
With the limitations of traditional surveillance systems, the higher cost, and the personnel requirement of commercial systems, our paper proposes a surveillance system that addresses the issues of cost, mobility, and efficiency using drones. Our method integrates computer vision to monitor aquaculture sites, such as fish feeding activities , inspecting nets, moorings, cages, and detecting suspicious objects (people, ships). These tasks require varying inspection facilities, adding difficulty to implementing a vision-based aquaculture surveillance system. To address this, we propose a low-cost and cloud-based autonomous drone equipped with a single camera to perform surveillance. The autonomous drone can capture the multiview video of the scene for inspection. The drone becomes an intelligent flying robot that captures distant objects and valuable data. These data will be processed using decision-making algorithms to provide information on how to optimize fish production and add security to the aquaculture site. The system has an aquaculture cloud, which is a private storage cloud that stores surveillance data for big data analysis. The cloud receives the inspection data from the drone to perform object-recognition activities. This process is faster since the cloud has a higher and more powerful computing ability as opposed to using drones with onboard computational power. This ability enables the drone to have faster speed and longer navigation time for its inspection activities. The cloud is also capable of understanding the scene, detecting suspicious activity, and detection. Artificial intelligence using computer vision provides a real-time monitoring capability for an aquaculture site. Computer vision offers an ideal automatic, noninvasive, economical, and efficient method to monitor facility-specific activities at an aquaculture site, in order to address the problem of expensive sensors . There are different types of object-detection and recognition models for the Artificial Intelligence (AI) services of our cloud system. However, conventional activity-recognition models cannot perform well on captured videos that are from a textureless water environment, especially for the evaluation of fish feeding intensity. To deal with this difficulty, we integrated our motion-estimation neural network to compute the optical flows from above-water images, since fish movements during feeding time produce optical flow. The use of optical flow enables the system to evaluate the intensity of fish feeding, and helps to detect excess feeding to help regulate the feeding process.
