Real TimeDetection and Tracking Method of Pilot’sHeadPositionBased on MTCNN-DeepSORT
Limin Zhang, Jie Jiang, Wei Fang, Kai Liu · Journal of Physics Conference Series · 2020
Abstract In view of the problems of the traditional head detection and tracking technology in pilot cabin simulation training, such as low precision, poor real-time performance and easy to be interfered by the outside world, we propose a head detection and tracking method based on MTCNN-DeepSORT combined with deep learning method. Firstly, the continuous video images in the cockpit are preprocessed by computer vision, and then the processed images are extracted by convolution neural network to detect the face. At the same time, the facial feature points are marked. Through the spatial coordinate calculation of the obtained facial feature points, the pilot’s head posture angle is obtained, and then combined with DeepSORT tracking algorithm can track the pilot’s head continuously and in real time, and finally complete the detection and tracking of pilot’s head position. The experimental results show that the method based on MTCNN-DeepSORT has higher detection accuracy, better real-time performance and stronger robustness than the traditional methods.