Mark-based vision for 3D vehicle tracking using least-squares and kalman filter

In this work, we analyse and compare two families of techniques for 3-dimensional tracking of vehicle movement using a fixed camera that provides vehicle images showing several, easily detectable, marks, fixed to the vehicle body. Algorithms are implemented in a mini-helicopter hover-stabilization application. The first family of techniques is based on non-linear dynamic least-squares (LS) algorithm for parameter estimation. The second family is based on optimal state estimation and extended Kalman filters (EKF). Both groups of techniques are first adapted to our problem and then discussed and compared from an analytical and numerical perspective