A simplified pupil location algorithm is proposed to address issues such as low precision, poor real-time performance, and susceptibility to environmental disturbances. Firstly, features are extracted from the input face image by the Dlib face model. Then, operations such as greyscale conversion, filtering, and erosion are applied to process the image of the critical areas of the human eye, and the initial positioning of the pupil center is achieved through the centroid method. Finally, a variable sliding window compensation algorithm based on differences in image gray values is used for coordinate correction. The algorithm comparison experiments are conducted on the BioID face dataset. The experimental results show that the average measurement time of each image in the dataset is only 26ms, which can meet the real-time requirements. Additionally, the pupil detection accuracy of the proposed algorithm is significantly improved compared with the traditional algorithm. The application experiments in a virtual driving environment show that the proposed algorithm can be applied to various demographic groups, has good real-time performance and robustness, which has important reference significance for the engineering application of eye-machine interaction.