Pig body weight prediction based on weight waveform graph

Livestock body weight and its changes are important indicators for evalua ting the growth, nutritional status, and health condition of livestock. The normal walking gait of livestock is a visual basis for assessing their health. Since the body weight waveform data generated by livestock freely walking on weighing platforms simultaneously records information on livestock body weight and walking gait characteristics, this study proposes to achieve dynamic weighing of pigs based on the pig weight waveform dataset. The paper first proposed an improved ResNet18 classification network model called ResNet18SV. Then, the ResNet18SV model is used for livestock gait classification, categorizing livestock gaits into four categories: fast, slow, normal, and linger based on their walking characteristics. Finally, a body weight regression model was tailored for each type of gait to accurately predict livestock weight. To validate the effectiveness and advancement of the proposed method, various classification models are used for gait classification comparative experiments, with the experimental results showing that the gait classification accuracy based on the ResNet18SV network is 97.13 %. Furthermore, based on the gait classification results, the pig body weight is predicted synchronously, with an overall average error of 0.28 %, close to the average error of static weighing (0.2 %). The experimental results demonstrate the good performance of our proposed method in livestock gait classification and dynamic weight prediction. By solely relying on inexpensive weighing platforms to obtain body weight waveform data, gait classification and weight prediction are achieved, thus offering a broader application value.