Conceptual Validation of High-Precision Fish Feeding Behavior Recognition Using Semantic Segmentation and Real-Time...

Area of Science:

• Aquaculture
• Computer Vision
• Machine Learning

Background:

• Unscientific aquaculture feeding practices lead to significant feed waste and water pollution.
• Current fish feeding behavior recognition methods struggle with complex environments, water interference, and target occlusion.
• Accurate recognition is crucial for optimizing feeding strategies and reducing operational costs.

Purpose of the Study:

• To develop an accurate and robust fish feeding behavior recognition method for intelligent aquaculture.
• To enable automatic bait casting machines to implement scientific feeding strategies.
• To reduce farming costs and improve aquaculture management efficiency.

Main Methods:

• Utilized semantic segmentation to accurately segment fish targets in images, overcoming complex background challenges.
• Developed a novel fish feeding behavior recognition model that analyzes aggregation characteristics during feeding.
• Input segmented fish images into the recognition model to identify feeding behaviors.

Main Results:

• The proposed method demonstrates excellent robustness and real-time performance in recognizing fish feeding behavior.
• Successfully handled complex water backgrounds and fish target occlusion.
• Provided an efficient and reliable solution for the aquaculture industry.

Conclusions:

• The developed method offers scientific support for intelligent aquaculture systems.
• It enhances aquaculture management and production efficiency through precise feeding behavior recognition.
• Future work may involve integrating multimodal data to further improve model robustness.