A bilateral symmetry based pose normalization framework applied to livestock body measurement in point clouds

Abstract

The shape of a livestock is a vital indicator of its health and value, whether for breeding or growth. Cutting-edge remote sensing technology provides an efficient and affordable way for acquiring point clouds of livestock, so that automated procedures of obtaining body measurements need to be established. A novel pose normalization method for 3D point clouds of livestock with similar forms of cows or pigs based on its bilateral symmetry properties is proposed in order to increase the degree of measuring automation. The proposed algorithm is combined in a hybrid scheme, which serves as the pose normalization procedure in an automatic body measurement system for livestock. Our extensive experiments on both synthetic and real world point clouds data show that the proposed approach has potential for generalizing well across livestock species and handles noise successfully in test data. In addition, the proposed pose normalization scheme outperforms current standard approach Principal Component Analysis (PCA) and state-of-the-art pose normalization method for pigs.