Spatiotemporal Analysis of “Jello Effect” in Drone Videos

A computational method for analyzing the “jello effect” caused by high-frequency vibrations of the camera with a rolling shutter mechanism is demonstrated. We use an optical flow model to analyze the “jello effect” in drone videos. Motion vectors and hue-saturation-value color map are used to visualize the spatiotemporal patterns. Our observations show that poor insulation between the drone structures and the camera creates a persistent “jello effect.” An impact of a projectile on the drone can cause a brisk “jello effect” where the magnitude of the shearing, rotational, and translational movements are proportional to the energy of the impact. By scaling the motion vectors in the optical flow, we are able to amplify the movements in the video to observe subtle distortions that are invisible to the naked eye. Our approach combines optical flow and visualization methods for real-time video analytics of drone footage. The study explores inferring physical interactions through non-ideal effects on digital systems. Applications include near real-time drone situational awareness.