In this study, an eye-tracking system based on FPGA hardware and the center of gravity algorithm is proposed. It captures video via LUPA300 high-speed CMOS and obtains the motion of eye pupil and bright spot through the continuous video frames by using a FPGA implemented the center of gravity algorithm. To obtain a reliable tracking accuracy, a series of binarization and mathematical morphology operations (such as corrosion and expansion) are adopted as a pre-procession step. The results show that FPGA can implement the center of gravity algorithm to track eye motion at high speed, the binarization and corrosion expansion pretreatment can improve the tracking accuracy effectively.