A Scene-Aware Augmented Reality Recommendation System
Hua Li, X. Liu, Salim Suedi Rashidi · Advances in transdisciplinary engineering · 2024
Augmented Reality (AR) technology offers users an unparalleled immersive interaction experience. In indoor scenarios devoid of GPS, AR-based recommendation technologies have yet to achieve effective real-time application. This paper introduces a scene-aware AR recommendation system that integrates cutting-edge instance segmentation and feature point matching technologies, alongside a versatile recommendation algorithm, to provide highly personalized AR content recommendations closely linked to the user’s immediate environment. Initially, the system utilizes an enhanced Mask R-CNN model for precise scene segmentation, followed by the extraction of key feature points combined with real-time scene information to devise an efficient recommendation algorithm. Experimental validation in real-world scenarios has demonstrated the system’s effectiveness in improving the accuracy of AR content recommendations and enhancing user satisfaction within indoor environments. Notably, the recommendation accuracy exceeds 80%, signifying a substantial advancement in AR technologies. This research demonstrates potential to profoundly enhance user interaction and satisfaction by providing more accurate and contextually relevant content recommendations in GPS-limited indoor environments.