2020 Volume 82 Issue 3 Pages 277-283
In this study, the terrain images captured in front of a tractor using a stereo-vision camera were reconstructed in three dimensions (3D). A stereo-vision camera was installed in various arrangements to verify the influence of its position and angle on accuracy of 3D-reconstruction. The obtained images were converted into 3D space information through image processing methods in using Matlab. The accuracy of the reconstructed ground was evaluated by comparison between actual and reconstructed ground height onto which left and right tractor wheels passed every 0.5 m. The root mean square error (RMSE) of the extracted tractor path’s height (between the reconstructed ground and the actual ground) was approximately 1.3-2.1 cm. The overall RMSE of the ground captured by the stereo-vision camera, which was tilted at 20° and mounted onto the upper side of the tractor weight, had the highest accuracy.