Smart Aquarium Caring System Using Image Processing and IoT

The expansion of the ornamental fish sector significantly contributes to Sri Lanka's economy. Nevertheless, manual practices within aquariums pose various limitations, impeding further growth. Manual methods for disease identification and growth detection heavily rely on expertise, while water quality monitoring practices are complex and costly. Accordingly, this research proposes a hybrid approach using image processing and IoT to automate the crucial aspects of aquarium care, including disease detection, water quality monitoring, and growth prediction. During disease detection, three different CNN architectures' performances were compared, and ResNet50 achieved the highest accuracy of 99.21%. The classifier identifies three types of fish diseases, including Redspot, Whitespot, and Tailrot. Water quality is monitored by measuring the pH, turbidity, and temperature levels in a tank using sensors in real time. The ESP32 microcontroller was used as the connector, and the Arduino IDE was utilized for the programming. Due to cost-effectiveness, the dissolved oxygen amount is predicted by the Random Forest Regressor with an R-squared of 0.83. Additionally, fish growth detection is performed to identify different growth stages of 1-6-month-old goldfish, and the ResNet50 achieved 85.5% accuracy. By integrating all three aspects, the proposed approach aims to provide a comprehensive and cost-effective solution for aquarium care.