A Lightweight Real-Time Face De-Identification Privacy-Preserving Algorithm Based on Airport Equipment

Aiming at the needs of passenger face privacy protection in the process of airport security check, combined with the lightweight characteristics of airport security equipment and the real-time nature of face detection, a lightweight real-time face de-identification privacy protection algorithm based on airport security equipment is proposed. The algorithm is divided into three parts: face detection, face tracking, and face de-identification. In the face detection stage, a lightweight real-time face detection based on the MediaPipe framework is proposed to detect the initial position of the face region using the real-time mobile application-optimized one-shot detector model SSD; in the face tracking stage, a bounding box is predicted from the coordinates of the face keypoints of the previous frame as the input of the current frame, which avoids the use of face region detection model on each frame, and a twin network based on spatiotemporal context information to solve the occlusion problem; in the face de- identification stage, StyleGAN is used to generate false face information. It is shown through experiments that the de- recognition security of this algorithm is higher.