correctly mapped by the HAR system. The initial position is determined by tapping to the representation of the real environment on the handheld device's display. A ray is cast from the device's camera to the tapped initial position. The object's position can then be adjusted along the epipolar line. SlidAR does not use virtual depth cues and it also enables the positioning of virtual objects in mid-air. We have also implemented another SLAMbased 3D positioning method called HoldAR, which is similar the device-centric method first introduced by Henrysson et al. [11]. In HoldAR, a virtual object can be freely positioned by fixing it's position to the handheld device and physically moving the device. Virtual depth cues are displayed on a ground plane.
We conducted a user experiment to evaluate the efficiency of SlidAR against HoldAR. We asked the participants to position virtual objects to the real environment. The results showed that SlidAR was significantly faster and required significantly less device movement. The subjective feedback on SlidAR was also rated significantly higher. Although not significant, we observed that the positioning accuracy was also higher when using SlidAR.
The main contribution of this paper is the SlidAR 3D positioning method for SLAM-based HAR systems. SlidAR does not require special hardware and it could be implemented to a vast variety of consumer handheld devices suitable for AR. Even though the method is developed for SLAM, it can be applied to marker-based HAR and it can be useful in any scenario where accurate 3D positioning of virtual objects is required. We proved the efficiency of SlidAR in a user experiment and we believe this is the first HAR 3D positioning experiment to have virtual objects being associated accurately to real world objects. The insights acquired from our experiment can be helpful in the design of future HAR systems and user experiments.
The rest of the paper is organized as follows: Related work is discussed in Section 2. Section 3 describes the details of the two positioning methods we used in the experiment. Sections 4 and 5 explain the design of our experiment and the results, respectively. Finally, the results are discussed in Section 6 and future work in Section 7.