Construction of statistical shape model of real cattle and its application to body measurement

Parameter Measurement of animals can be used to characterize animals and is important for growth monitoring and assessment of animal welfare. Consumer-grade point cloud data acquisition equipment is inexpensive, has fast imaging speed, and is accurate enough for agricultural applications. There are many works related to the application of 3D point cloud data processing techniques to animal body measurement. However, there are still many challenges in applying these works to actual production, such as serious missing point clouds caused by occlusion and unstable positioning of keypoints for body measurement. In this study, a statistical shape model of real cattle is constructed on a topologically consistent 3D surface dataset. On this basis, the 3D mesh reconstruction algorithm and body measurement technique based on low quality point cloud are studied to overcome the application problem of animal body measurement directly on low-quality point cloud data. We validate the method using low-quality point cloud data of 99 cattle. The article calculated the four indicators of chest depth, ilium width, oblique body length and heart girth, and the overall measurement accuracy reached 91.40%. Compared with the latest research results, it shows that the method has strong robustness and feasibility for livestock body measurement.