3D reconstruction of chicken based on color and thermal images

Abstract

<b>Abstract.</b> The body parameters and surface temperature are both important indicators evaluating the health and welfare status of chicken. If we can both monitor body parameters and surface temperature of chickens, it will be intuitive to observe and judge their health status. The objective of this study was to propose a 3D reconstruction method of chicken fusing the color and thermal information. The binocular color cameras and a thermal camera were utilized to acquire images. The image registration between the left color image and the thermal image was achieved by adjusting the scale and position, as well as adopting projection transformation. Then the PatchMatchStereo (PMS) algorithm was utilized to predict the depth map according to binocular color images. Based on the registered color image with depth map, as well as the registered thermal image with depth map, the scene color point clouds and scene thermal point clouds can be reconstructed, respectively. Then the target point clouds of both color and thermal information were extracted from the scene point clouds based on known distance information and DBSCAN algorithm. The MAE and RMSE values of the 3D reconstruction method were 8.95% and 9.35%, respectively. The results showed that this approach can successfully reconstruct the complete point clouds of the chicken with color and thermal information, which promoted the application of intelligent monitoring technology in animal behavior research.