Continuous tracking of moving objects in long-distance imaging through a turbulent medium using a 3D point cloud analysis

Abstract

Preserving object detections and maintaining continuous, uninterrupted tracking routes of moving objects is a difficult challenge in long-distance imaging through a turbulent medium that extremely degrades the video. We introduce a new method for the continuous tracking of moving objects in such conditions. First, a number of steps are applied, including global video stabilization, reduction of local oscillations, and finding spatio-temporal “suspicious sections” in the video that may contain moving objects. This phase produces a 3D point cloud that represents suspicious moving objects. Subsequently, the possible routes in the cloud are analyzed, producing the final routes. A comparison to the current standard shows an improvement in the object tracking stability.