Search

In aquaculture, quantifying the real-time feeding behaviour of fish is essential for making feeding decisions. However, most existing methods for assessing fish appetite are inefficient and subjective. To address these issues, this study proposes an improved tilapia feeding level classification model with ResNet34. First, we introduce the attention module CBAM into the ResNet34 model to adjust the attention of the model according to the importance of different channel features and enhance the effective extraction of important features. We then used migration learning to transfer the knowledge learned from the source data (ImageNet dataset) to the tilapia ingestion image dataset, which allowed us to train the tilapia ingestion behaviour classification model faster while retaining the pre-trained model. Experimental results showed that the improved ResNet34 model in this study achieved an accuracy of 99.72%, an improvement of 7.84 percentage points over the original model. In addition, the model outperformed models such as MobileNetV2, AlexNet, VGG11, ShuffleNet_v2_x0_5 and ResNet18 in terms of accuracy, precision, recall and F1 scores.These results suggest that the proposed method can accurately identify feeding behavior of fish and provide a scientific basis for determining feeding amounts.

Pigeon is a kind of poultry with high practicability and economic value, and humidity is an important indicator of the intensive breeding of pigeons, which is closely related to the healthy growth of pigeons. Humidity changes are complex, dynamic, and nonlinear. Accurately predicting humidity values and analyzing their changing trends is crucial to the growth of pigeons. Addressing low prediction accuracy and poor generalization of traditional humidity prediction methods, this paper proposes a combination based on partial least squares (PLS) data dimensionality reduction, Savitzky-Golay (SG) filter, and sparrow optimization algorithm (SSA) Predictive models to improve humidity forecast accuracy. First, the SG filter is used for data smoothing to remove abnormal noise in the data signal to enhance the feature learning ability of the prediction system. Next, the original data is selected through the PLS method to select information higher than the set threshold. Valuable information features are extracted for data dimensionality reduction. In addition, SSA is used to optimize key parameters and structure of the BiGRU model for improved prediction. Finally, a combined prediction model based on SG-PLS-SSA-BiGRU was built to simulate and predict the humidity of the pigeon house. The experimental results demonstrate the proposed model’s high accuracy and stability with minimal prediction error fluctuations, fast calculation speed, good feature extraction ability and generalization ability, and is very effective and reliable for predicting pigeon house breeding humidity.

Dissolved oxygen content is a key indicator of water quality in aquaculture environment. Because of its nonlinearity, dynamics, and complexity, which makes traditional methods face challenges in the accuracy and speed of dissolved oxygen content prediction. As a solution to these issues, this study introduces a hybrid model consisting of the Light Gradient Boosting Machine (LightGBM) and the Bidirectional Simple Recurrent Unit (BiSRU). Firstly, Linear interpolation and smoothing were used to identify significant parameters. LightGBM algorithm then determines the significance of dissolved oxygen by eliminating irrelevant variables and predicting dissolved oxygen in intensive aquaculture. Finally, the attention method was implemented to map the weighting and learning parameter matrices, so enabling the BiSRU’s hidden states to be assigned different weights. The findings shown that the presented prediction model can accurately anticipate the fluctuating trend of dissolved oxygen over a 10-day period in just 122 seconds, and the accuracy rate reached 96.28%. Comparing the model effects of LightGBM-BiSRU, LightGBM - GRU, LightGBM-LSTM, and BiSRU - Attention takes the least time. Its higher prediction accuracy can provide an essential reference for intensive aquaculture water quality regulation.

Accurate counting is difficult in the case of large numbers of overlapping and adhering fry. In this study, we propose a lightweight target detection counting method based on deep learning methods that can meet the deployment requirements of edge computing device for automatic fry counting while obtaining a high counting accuracy. We improve the structure of YOLOv4-tiny by embedding different attention mechanisms in the cross stage partial connections blocks of the backbone network to enhance the feature extraction performance. In addition, the low efficiency of feature fusion in the original model is also addressed by adding different attention mechanisms to the neck network structure to promote the effective fusion of deep feature information with shallow feature information and improve the counting accuracy. The experimental results showed that the six models proposed in this study improved the model accuracy and recall to varying degrees compared with the original YOLOv4-tiny model, while retaining the advantages of the YOLOv4-tiny model in terms of its small number of parameters and fast inference rate. It was also shown that the CBAM(n)-YOLOv4-tiny model obtained by adding the CBAM to the neck network showed the most significant improvement, with an mean average precision (mAP) of 94.45% and a recall of 93.93%. Compared with the YOLOv4-tiny model, there were increases of 27.06% in accuracy, 30.66% in recall, 38.27% in mAP, and 28.77% in the F1-score, along with a 67.82% decrease in LAMR.