Path Reconstruction from Video Sequences Using Nadiral View and Optical Flow

In this article we propose an approach to reconstruct the path traveled by a human in a bi-dimensional horizontal space, based on information acquired by the sensors (video camera, gyroscope) of a hand-held mobile device. The input data consists of the readings from the sensors, i.e. the video signal recorded along with the readings from the gyroscope. The readings are synchronized, so that for each video frame we know the device orientation. The output is the trajectory of the mobile device's camera in the horizontal plane, as reconstructed based on the nadiral view and the optical flow, computed based on the video and motion sensor data. The reconstructed path is the traveled line in a bi-dimensional, horizontal plane.