This thesis presents a vision-based navigation for an indoor UAV utilizing only low cost cameras installed indoor. To overcome the limitation of the outside flight experiment, an indoor test-bed using multi-camera system is developed, which consists of four major components: the multi-camera system, the ground computer, the onboard color marker, and the quad-rotor UAV. The dynamic model of the UAV is the six-degrees-of-freedom nonlinear equations derived from Newton``s Law. The measurements are the visual information of the color marker attached to the UAV which is obtained periodically from multi-camera via computer vision algorithm. The extended Kalman filter considering the delayed measurement is designed to obtain the full 6 DOF pose estimation for the UAV. The quad-rotor UAV is considered as a platform vehicle since it has simple dynamics and can be effectively operated in narrow indoor environments. The control system is designed based on the classical PID control. This thesis finishes with several experimental results illustrating the performance and properties of the proposed the indoor test-bed and the vision-based navigation algorithm.