Object shape estimation and tracking based on earth mover's distance and image moments

Object tracking and its shape estimation from a variety of different measurement sensors is an important problem in many autonomy, robotics, and aerospace system applications. This dissertation proposes tracking algorithms considering two types of measurements: dense illuminance-based measurements and sparse measurement points. Visual tracking algorithms are based on dense illuminance-based measurements, which use the silhouette, color and texture information to represent and track the target. However, when measurements are sparse and only have the position information, extended object tracking (EOT) algorithms are used to solve this problem. For visual tracking using dense illuminance-based measurements, an efficient iterative Earth Mover's Distance (iEMD) algorithm is proposed. The Earth Mover's Distance (EMD) is used as a similarity measure to search for the optimal template candidates in feature-spatial space in a video sequence. Local sparse representation is used as the appearance model for the iEMD tracker. For extended object tracking using sparse measurement points, two models are proposed in this dissertation considering the geometric characteristics of the targets. An image moment-based model is proposed for tracking and estimating the position and shape of the rigid extended object. To track and estimate the shape of the elongated deformable object, a B-spline chained multiple random matrices representation is proposed. Simulation and experimental results are presented to validate the proposed tracking algorithms.