Determining the motion of a remotely piloted vehicle from a sequence of images

The author addresses the problem of determining the motion of an airborne passive imaging sensor, e.g. images acquired by a remotely piloted vehicle. The solution has applications in passive navigation. From a number of point correspondences from two or more brightness images, the relative motion of the sensor and the depths of the imaged scene points can be computed up to a global scale factor. Assumptions are a static scene, a known camera focal length, and that the correspondence problem has been solved. The equations that describe the motion and scene structure are nonlinear in the camera model parameters. The camera model is based on the collinearity condition for perspective projection. The approach differs in choice of model parameters and in formulation of the collinearity condition. The solution is based on Newton´s method for nonlinear parameter estimation. The Levenberg-Marquardt extension for discrete-least-squares is applied to the overdetermined system to improve the algorithm´s range of convergence and robustness with respect to noisy image measurements. Monte Carlo simulations demonstrate the improvements in the algorithm