Answers are grounded in this paper and cite their sources.
In aquaculture, accurate assessment of fish appetite is crucial for developing effective feeding strategies. However, variations in lighting conditions in outdoor ponds can lead to unstable image quality, affecting subsequent processing. To address this challenge, this paper proposes a novel method for evaluating fish appetite in outdoor pond environments. First, this study used V-channel contrast-limited adaptive histogram equalization (V-CLAHE) for image contrast enhancement, highlighting fish feeding behaviors and water surface fluctuations. Next, based on the feature-enhanced kinetic energy model (FEKEM), fish feeding behavior information was extracted and quantified. Finally, the fish TCN (FTCN) model was employed to determine the appetite levels of fish. FTCN is a novel temporal classification model developed by improving the modern temporal convolutional network (ModernTCN), incorporating efficient channel attention (ECA), and utilizing a multilayer perceptron (MLP). The combination of the improved ModernTCN and ECA is leveraged for the efficient extraction of temporal features. The modified ModernTCN improves the modeling capability for complex time-series data, while ECA captures inter-channel dependencies, thereby enhancing the focus on key features. An MLP is utilized to capture deeper feature representations and achieve precise classification through the softmax function. In outdoor pond scenarios with significant illumination variations, a dataset was constructed based on fish feeding behavior time-series data extracted using FEKEM. The performance of the proposed FTCN model, as well as the combined method of FEKEM and FTCN, was systematically evaluated. Experimental results show that FTCN achieves a classification accuracy of 98.13 %, outperforming existing methods such as TimesNet (95.82 %), Nonstationary_Transformer (95.93 %), and Informer (94.61 %). This effectively validates its superiority in temporal classification under complex environmental conditions. Moreover, the combined method of FEKEM and FTCN demonstrates superior performance in fish appetite assessment compared to existing approaches (with the highest accuracy at 92.86 %), further confirming the effectiveness and practicality of the proposed approach in fish feeding behavior recognition and appetite evaluation.
No takes yet. Share an insight, caveat, or question.
A 2025 study studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: