Multi-User Identification and 3D Positioning Using Multi-View Images via Contrastive Learning

Recently, computer vision (CV)-aided positioning techniques have emerged as a promising approach to achieve precise positioning in wireless communications. However, due to the occlusions caused by obstacles, to provide a constant view of the user equipment (UE) is a challenging task. In this paper, we propose a novel multi-view positioning scheme, called contrastive learning-based multi-user identification and positioning using multi-view images (C-IPM), for the multi-user communication systems. To be specific, we first find out the positions of bounding boxes for objects and determine the class of the detected objects. Based on visual similarity of the images captured at different viewpoints, we identify the same UE across multi-view images. Once the UE is identified, we extract the 3D position corresponding to the UE's pixel in the images. From the simulations, we demonstrate that the proposed C-IPM scheme outperforms conventional positioning techniques in terms of positioning error.