Abstract: Eye gaze targeting technology based on eye movement tracking can free target tracking and aiming control of unmanned weapons from the need for limbs and make it possible to "attack whatever is seen", which is an important control mode of unmanned weapons in the future. A mixed weighted feature method was proposed in this paper. Human eye region detection was obtained through Gabor operator filtering, integral graph calculation, introducing local region variance as weight to weight mixed feature codes and combining it with cascade classifier training. The experimental results show that the method in this paper is better than the commonly used Haar-like and LDP methods, and the false detection rate shows a downtrend with the increase of series. This method can enhance the detection rate of human eyes and reduce false detection rate, providing a possible technical way to meet the requirements of real-time and accuracy of unmanned weapon eye gaze targeting.