This paper presents the design and implementation of a visual servo system, which uses a stereo machine vision system to guide an omnidirectional mobile robot for real-time object tracking. The vision system consists of two image sensors so that the position of the target in the space can be obtained. The real-time vision system is placed on the omnidirectional mobile robot for navigation. The tracking control of the robot is done by feedback linearization combined with PID control. The tracking controller is implemented on a digital signal processor. For the machine vision system, the image processing algorithms are pipelined and implemented on a field programmable gate array (FPGA) device to meet real-time constraints. The effectiveness of the tracking controller and stereo machine vision system is verified through simulation and experimental studies.