Augmented Reality Area Recognition Method for Industrial Scenarios

At present, the industrial scenarios may have multiple similar targets in the camera field of view at the same time. Aiming at the existing augmented reality methods that cannot distinguish these targets, the augmented reality area recognition method is proposed. Firstly, the depth camera is used for completing the 3D registration of the factory area and constructing the area targets that include natural features as well as background features. Secondly, a dual map reloading strategy is designed to realize the position setting of the virtual object and cross-platform visualization. Finally, the robust relocation algorithm combined with the pose graph and bag-of-words, is used to implement the identification of the area target. The experimental results tested in the TUM dataset and real scene show that the average value of root-mean-square absolute trajectory error is 0.076, the frame rate exceeds 60, and multiple similar targets under the same field of view are distinguished. The method has high accuracy and good real-time performance, can effectively expand the application scenarios of augmented reality.