Iterative offset-based method for reconstructing a mesh model from the point cloud of a pig

Abstract

In the livestock industry, assuring the quality of livestock products is as important as reducing breeding costs. The quality of livestock is generally evaluated based on their body form and weight. More specifically, a series of measurements using depth cameras or laser scanners are conducted to obtain 3D data in the form of a point cloud, in an attempt to monitor and determine the breeding status of livestock. In previous studies on livestock weight estimation, major body dimensions were first calculated using point cloud data and based on the results, livestock weight was estimated using weight estimation formulas. These approaches, however, failed to consider the entire body shape of the livestock over the course of weight estimation. In the present study, a mesh reconstruction method for the point clouds of pigs, and especially those with a large amount of noise or missing data points, was proposed as a means of pig weight estimation. The basic principles of the mesh-model reconstruction process are as follows. First, a primitive shape is created in the form of a mesh and the travel distance for each node contained in the mesh model is calculated. The nodes are moved according to the calculation, and triangular element division and smoothing are performed for quality improvement. Iterations of node travel and quality improvement processes are conducted to reshape the mesh model to fit the shape of the corresponding point cloud. To verify the proposed method, point clouds were obtained by scanning actual pigs and mesh models were automatically reconstructed to fit the obtained point clouds. The results showed that the maximum distance between the point cloud with noise removed and the corresponding data points of the mesh model, was within 3.12 mm for all cases. These results confirmed that the reconstructed mesh model provided a satisfactory performance.