A reconstruction method for incomplete pig point clouds based on stepwise hole filling and its applications

Abstract

The 3D model accurately depicts the surface characteristics of pigs, enabling measurement of their body size and prediction of the weight. However, multi-view 3D point cloud reconstructions of pigs often suffer from significant missing areas in leg and torso regions due to factors like railing obstructions and camera blind spots. To address this issue, this paper proposes a method for reconstructing incomplete pig point clouds based on stepwise hole filling. This approach converts the point cloud into mesh, initially filling part of the large, high-curvature holes that are difficult to handle based on pig morphology to narrow their extent, followed by filling remaining areas. Experimental results show that the completion effect of this method is visually superior to existing completion methods. The mean relative errors for calculating cannon bone girth, chest girth, and abdominal girth using the completed model compared to manual measurements were 5.04 %, 3.83 %, and 3.51 %, respectively, representing reductions of 1.24 %, 11.47 %, and 9.48 % compared to the method of directly using incomplete point clouds. In addition, utilizing the watertight properties of the mesh model completed by this method, the volume of the pig was calculated, and a volume-based Logistic regression weight estimation model was established, achieving a mean absolute percentage error (MAPE) of 4.06 %. This underscores its high precision in estimating pig weight.