Concept: Smart Fish Farming proposes a digital ecosystem that integrates AI, machine learning, and image processing technologies to monitor fish in their natural aquaculture environment. The system employs underwater cameras, sonar, and stereo vision technologies to capture detailed images and data on fish behavior, size, and health indicators. Advanced AI algorithms then analyze this data to generate actionable insights, enabling farmers to optimize feeding, predict harvest times, and detect diseases early.
Additionally, a consumer-facing mobile application offers product transparency by analyzing fish quality based on captured images.
Methodology: The methodology of Smart Fish Farming is divided into several key stages:
- Data Collection: A network of low-cost, high-resolution cameras captures real-time images of fish in underwater environments.
- Image Preprocessing: Advanced image processing techniques enhance underwater images by normalizing light, reducing noise, and segmenting fish from the background.
- AI and ML Analysis: Trained AI models classify fish based on species, orientation, and size, while detecting any abnormalities. Fish tracking allows the system to analyze behavior, enabling stress or health issues to be identified early.
- 3D Shape and Landmark Annotation: The project adapts techniques from facial recognition (such as Ensemble of Regression Trees) to fish morphology. This allows for precise dimension measurement, weight estimation, and shape analysis.
- Hardware Acceleration: Processing multiple images simultaneously requires high computational power. To achieve this, the system incorporates FPGA hardware accelerators, which enable real-time analysis without compromising accuracy.
Consumer Mobile Application: Selected capabilities from the fish farming system are also available on a mobile app, which allows consumers to assess fish quality and freshness by analyzing specific fish features and verifying their origin and rearing conditions.