In this paper, a new pose estimation solution for perspective vision system is presented and applied to a civil aircraft landing phase. Vision sensor is used to overcome the need for external technologies and runway knowledge. Two nonlinear observers are proposed for on-line estimation of the deviations of the aircraft w.r.t. the runway. These nonlinear observers are based, first, on a high-gain approach, and then on a high-order sliding-mode approach, allowing robustness and finite time convergence. The originality of the presented results consists in estimating deviations with redundant informations in order both to increase efficiency of the observer and guarantee observability. Simulation results obtained on a realistic landing scenario are presented.