Laplacian Vision

Yuta Itoh, Jason Orlosky, Kiyoshi Kiyokawa, Gudrun Klinker
2016 Proceedings of the 7th Augmented Human International Conference 2016 on - AH '16  
Figure 1 : (a) Our Laplacian Vision system with an Optical See-Through Head-Mounted Display (OST-HMD) and user-view camera placed behind the display screen. (b) A miniature catapult launches a target ball to a landing pad. (c) A scene with see-through visualization of a predicted ball trajectory, which is superimposed onto the view in realtime. (d) A magnified view through the system of the landing area from the side. ABSTRACT Naïve physics [7], or folk physics, is our ability to understand
more » ... ical phenomena. We regularly use this ability in life to avoid collisions in traffic, follow a tennis ball and time the return shot, or while working in dynamic industrial settings. Though this skill improves with practice, it is still imperfect, which leads to mistakes and misjudgments for time intensive tasks. People still often miss a tennis shot, which might cause them to lose the match, or fail to avoid a car or pedestrian, which can lead to injury or even death. As a step towards reducing these errors in human judgement, we present Laplacian Vision (LV), a vision augmentation system which assists the human ability to predict future trajectory information. By tracking real world objects and estimating their trajectories, we can improve a users's prediction of the landing spot of a ball or the path of an oncoming car. We have designed a system that can track a flying ball in real time, predict its future trajectory, and visualize it in the user's field of view. The system is also calibrated to account for end-to-end delays so that the trajectory appears to emanate forward from the moving object. We also conduct a user study where 29 subjects predict an object's landing spot, and show that prediction accuracy improves 3 fold using LV.
doi:10.1145/2875194.2875227 dblp:conf/aughuman/ItohOKK16 fatcat:lklgx2edf5hltcjbqp2owwzeau