Detection of Porcine Huddling Behaviour Based on Improved Multi-view SSD

The huddling behaviour of pigs is closely related to disease prevention and comfort degree. In order to detect the huddling behaviour of pigs in complex breeding environment, a detection method for huddling behaviour based on improved multi-view Signal Shot MultiBox Detector (SSD) is proposed in this paper. Firstly, in order to solve the deficiency of SSD for small object detection and close-contact object detection, the improved multi-view SSD is developed to detect pigs in complex breeding environment. Meanwhile, MobileNet_v2, focal loss function and network parameters transfer are applied in this paper to further improve the detection accuracy and speed of SSD. Secondly, the dispersion of pigs is proposed to signify the aggregation degree of pigs. Finally, the threshold of dispersion is set to detect the huddling behaviour of pigs. The experimental results show that the huddling behaviour could be detected with an accuracy of 94.5%, a precision of 94.9%, a recall of 94.0%. The results indicate that this method can be used to detect huddling behaviour of pigs.