Efficient Structure from Motion for 3d Reconstruction of Static Background and Moving Targets

Modern munitions need the ability to track moving targets with passive sensors. This requires a 3D reconstruction of both the background scene and moving objects. Normally this is done with stereo vision to generate a 3-D scene. Unfortunately, the use of paired cameras on munitions is difficult due to the need for spatial separation of the cameras for resolution and the general desire to minimize the weight and cost of seeker payloads. Single cameras provide 2-D representations that are intrinsically ambiguous, since one dimension is lost in the projection from the 3-D world onto a 2-D imaging sensor. One way to overcome this loss is to use the dimension of time and movement of the camera. Given a series of images of a scene, typically taken by a video camera, it is sometimes possible to recover some of this lost 3-dimensional information with a technique called Structure from Motion. Robotic Research, LLC in conjunction with the Jet Propulsion Laboratory proposes to develop an efficient hardware-based Structure from Motion (SFM) module for static scenes and moving targets. This research effort will advance the ability to jointly reconstruct static backgrounds and moving targets using single passive sensor.