An Automatic Measurement Method of Animal Body Size Based on Contour Segmentation

In the process of animal husbandry production, measuring the body size of livestock is an important part, because the body size of livestock is one of the important evaluation indicators for examining the breeding performance and development status of breeds. The traditional manual measurement method not only has a large workload and the measurement results are greatly affected by human subjectivity, but also easily causes stress reactions in livestock. With the rapid advancement of computer technology, computer vision has been extensively applied in animal husbandry production. Based on the practical problem of livestock body size measurement in animal husbandry production, an automatic measurement method of animal body size based on contour segmentation is proposed in this paper. This method utilizes the DeepSnake algorithm to segment target contours in two-dimensional images and extract skeletal structures to identify key points for body size measurement. Finally, by integrating point cloud data obtained from a depth camera, contactless automatic measurement is achieved. At the same time, we propose a standard standing posture classification method based on skeleton features. In our experiments, 10,189 images are used for training and 667 images are used for testing. The silhouette skeletons segmented by DeepSnake on the test set are utilized to train an SVM for classifying standard standing postures. Finally, the standard standing posture classification and body measurement experiments were conducted on a short video. The accuracy of standard standing posture classification is up to 88.9%, and the average accuracy of body measurement reached 92.7%. The results show that this method has broad application prospects.