Motion vision based structure estimation in forest environment

Motion vision can be used to determine world structure from a video sequence. In harvester machine automation, the potential is that trees could be measured from a distance. Based on the measurements, tree cutting could be optimized and harvester automation increased, resulting in higher resource utilization efficiency. However, a natural environment poses challenges to any computer vision task. This paper presents computer vision algorithms that are applied to a forest environment. The results show that dense optical flow can be computed from a real-world forest data accurately enough as to enable instantaneous dense structure estimates of the visible image scene.