A Fatigue Detection system has been developed using non-intrusive vision based approach. The system uses a Logitech USB camera which points towards driver's face and monitors face and eyes to detect driver fatigue. The system is developed on Linux operating system and used DM3730 processor as hardware. The algorithm is developed to estimate whether eyes are open or closed and fatigue is estimated using PERCLOS method. Normal human blinks eyes 12 times in a minute. If the eyes are closed for 15 consecutive frames in a minute or if PERCLOS > 80% than system issues warning to stop the vehicle. The algorithm is tested on 45 different persons i.e., 15 women, 15 men and 15 persons wearing spectacles and the detection rate is 99.2%. The system takes <5ms of time to detect whether eyes are open or closed and the hardware used is small in size and easily implementable.