Page 105 - Kaleidoscope Academic Conference Proceedings 2021

Connecting physical and virtual worlds

and unfolding so that it avoids the need for the user to
Algorithm 2: Dimension_Scanning manually go to the store and try it out.
Input first and second Hitresults anchor from AR
Raycasting.
Create an anchor as hit results does not happen in the same
frame
startAnchor =
session.addAnchor(hitResult.getHitPose());
Obtain the location of first and second anchors
Pose startPose = startAnchor.getPose();
Pose endPose = hitResult.getHitPose();
Clean up the anchor since the tracking gets updated
session.removeAnchors(Collections.singleton(startAnch
or);
startAnchor = null;
Compute the difference vector between the two hit (a) Bed 1 (b) Bed 2
locations
dx = startPose.tx() - endPose.tx();
dy = startPose.ty() - endPose.ty();
dz = startPose.tz() - endPose.tz();
Compute the straight-line distance.
distance = (float) Math.sqrt(dx*dx + dy*dy + dz*dz);
return distance

(c) Cupboard (d) Chair

Figure 4 – Sample Home Interior Models in real
world view
(a) Vertical Plane (b) Horizontal Plane

Figure 3 – Plane detection

4.1 Spawning virtual models in real scene

Once the plane is detected, the 3D model to be rendered is
instantiated and deployed in the real scene. The Lean API
handles different transformations and gestures as shown in
Figure 4 where the white icon depicts that the 3D model can
be captured as an image while the red dot indicates the
recording action of the interior models while they are stacked
in the scene which are stored in the native gallery of the
mobile device. Transformations (scaling and rotation)
carried out on the models are shown in Figure 5. (a) Scale (b) Rotate

4.2 Realistic 3D view of the featured models Figure 5 – 3D transformations

It is not enough if users can only view the item from one
angle, thus the user should be able to calibrate the 3D model
with the information given to place the item correctly. Users
can also move around the interiors to get a clearer picture on
how it would look like from multiple perspectives. Figure 6
shows how a model would look in real scene during folding

– 43 –