AR in a Large Area Through Instance Recognition with Hybrid Sensors

This paper presents the concept of instance recognition with access points and vision for realizing augmented reality in a large indoor environment. The proposed study aims at contributing to the reduction of computational cost and mismatches that occur when the database size for instance recognition is large and includes similar textures. The proposed method consists of database construction and instance recognition through the database. The database construction process involves structuring pairs of images and access points for managing images through the access points. The objective of the instance recognition process is to find the best fit by incorporating access points and vision. The evaluation results show that the proposed method consumes 68% lesser computational time and has 10% greater recognition accuracy than our previous work using only vision.