Abstract:In order to overcome the problems of manual livestock body size measurement which is labor intensive, time consuming and stressful to the animals, a livestock body size measurement method based on the mirror reflection of single angle point cloud was proposed. A single depth camera Xtion was utilized to capture the sequences of point clouds of the scene, including pig and point clouds were stored in the computer, and then the scene point cloud, including the pig that bending degree of its back was not large was selected manually. Subsequently, the ground of the scene was deleted by using random sample consensus, the remaining scene point cloud was segmented based on Euclidean clustering, and then all the clusters were sorted according to the distance from the center of cluster to the coordinate origin, and the smallest one was extracted as the target pig. Pose normalization of the extracted pig was obtained by using a PCA based approach. Then, the pig symmetry plane was detected from the single angle point cloud by using the prior information of pig after pose normalization. Using the symmetry plane, the point clouds data of whole body surface of pig was obtained. Finally, by using the software developed by the institution of the authors, body size measurement was performed manually. The results showed that the average relative errors were 5.00% for body length measurement, 7.40% for hip width measurement and 5.74% for hip height measurement. This method offered a new way to measure the livestock body size.