Page 102 - Kaleidoscope Academic Conference Proceedings 2021

2021 ITU Kaleidoscope Academic Conference

need for object tracking methodology. To apply AR to an 2. PROPOSED SYSTEM
application of furniture shopping in an interior design project
to customers, the depth information needs to be captured to The architecture of the proposed system is shown in Figure
estimate the supporting surfaces like floor and then area, 1. In this system, input is obtained from the camera of the
shape and dimension of the space to be used can be handheld (mobile) device. Then, horizontal and vertical
ascertained. The AR solution needs to automatically arrange plane detection is deployed by means of anchors and feature
the selected furniture in a defined 3D space which requires points. Anchors are used to hold the virtual objects in place
linking spatial and functional relationships in app design [8]. when it is spawned in the real world. Anchors use the 3D
world coordinate system to map the location of the object
Another design issue in home furnishing is to ensure that the while the user is moving. When the object is spawned, it is
selected interior has the correct dimensions for the target given a value based on its position and orientation called
room, where a user may try multiple spatial configurations. pose. So, when anchors are created, the pose of the object
To align the virtual object in the targeted real-world 3D space shows the position and orientation relative to the world space.
in real time considering physical dimensionality constraints, Anchors use the concept of feature points to stay anchored to
it requires implementing functions to detect and monitor the place where they are spawned. Feature points are the
parameters like change in position, rotation, and scaling distinctive features in each frame of the system. These
dynamically in a real-world scenario. The latency becomes a features include different planes, textures, and colors.
critical factor not only in the user’s experience, but also in Marker-less tracking is deployed via SLAM [15] which uses
system performance considering available limited natural features of the physical environment to detect
computational resources in handheld devices. In [9], the position and orientation and increases mobility.
authors implemented “snapping to reality” as a proof-of-
concept prototype solution and claimed a faster response 3D registration identifies the features and locates the target
when assisted by snapping to dynamically extracted interior models with the help of video input acquired by the
constraints. However, their system requires special hardware, camera of the mobile device. Home interior modeling,
namely Microsoft Kinect 2.0 sensor, and desktop PC with texturing, lighting, rigging, animation, and dynamics are
NVIDIA GeForce GPU to implement. performed using Maya and imported into Unity for
establishing 3D projection of the interiors. By identifying
The proposed system utilizes marker-less augmented reality and tracing the surface area, projection models of the virtual
implemented on a mobile device for enhancing user objects are established and occlusion, if encountered, is
experience in applications of the interior home design related handled. The dimension scanning is performed to measure
with furniture shopping, with the system design goal of the size of the home interior model. Transformations such as
enhancing performance in terms of latency while meeting the scaling, color and texture change, rotation, drag, pinch, tap
user’s requirements. This requires support of the field-of- and other gestures can be applied to the model in a real-world
view and perspective of the user, thereby allowing the view via Lean API. It consists of parameters such as index
environment to react accordingly. The existing popular trial- (hardware id of the finger), age (how long the finger has been
ware application software is used design, experiment, and active in seconds), tap (whether the finger tapped the screen)
implement the proposed system. The Autodesk Maya [10] is and its corresponding tap count, pressure (current pressure of
deployed to model, animate, and render 3D scenes, while the finger), and screen positions. The touch heatmaps provide
Substance Painter [11] is used to texture and render the 3D an aggregate of an application’s user experience and interface,
meshes for assets that have a specific UV layout. By indicating the user’s focus as well as interest on every screen.
combining the AR Foundation framework with Lean Touch The 3D object models are established based on its vendors
in Unity [12], and dynamically handling the occlusion and input. These home interior design models are then captured
instant tracking of 3D feature points using the SLAM as an image or as a video clip and are stored in a native
technique, the deployment time of the home interior models gallery of the handheld device.
gets minimized. The system models have been developed to
meet requirements listed in ITU-T G.1035 “Influencing 3. AR MODEL AND ALGORITHMS
factors on quality of experience for virtual and augmented
Reality services” [13] and Q.4066 “Testing procedures of 3.1 Model rendering for marker-less tracking
augmented reality applications” [14]. The system model
described in this paper can be a candidate technology of The scenes where the 3D models are projected are captured
virtual and augmented reality services being standardized in using the device camera. After the scene is captured, the
ITU-T Study Group 16. ground plane on which the model should be rendered is
detected. The initial coordinate system is set up where the
The remainder of the paper is organized as follows. Section model should be rendered. This forms the ground plane of
2 provides the architectural details of the proposed system and the scene. Features around the ground plane include points,
Section 3 describes the AR model and development of lines, circles, and distances. Distances between these
algorithms. The implementation environment, experimental features and the depth of each feature point from the camera
results, and performance evaluations of the proposed system
are presented in Section 4, while Section 5 concludes the
paper with the scope for future work.

– 40 –