Page 103 - Kaleidoscope Academic Conference Proceedings 2021

Connecting physical and virtual worlds

Camera of Establishing 3D Object
Video input Handheld Device model

Making 3D Animation
Plane Detection Dimension
Scanning

Import model and
Marker-less animation data into
Tracking
program

3D Registration, Obtain
Camera Position
Object
Transformation
Gesture

Scale Color Recognition
Stack virtual
Establish Projection model on the real Occlusion

Model of virtual object scene Handling Rotate Heat Map

Texture

Figure 1 - Architecture of the proposed system
position are determined. Instant tracking involves matching component holds the plane’s boundary points which always
the calculated values of the initial frame with values forms a convex shape to the plane’s normal. As these are
calculated from each of the next frame as shown in Figure 2. updated, the mesh visualizer updates the visual
When the features do not match up, tracking is lost. The representation to perform horizontal and vertical plane
models are rendered within the view of the mobile tracker. detection. At that time, the camera module (AR Camera) is
Hence the application follows MVC architecture. Each scene detected and the trackable object which is to be rendered is
contains all models of that particular level, and the first scene
is the model selection. Once the model is selected, the scene
changes as per the tracking requirement. Initial Frame (frame 0)
Setup Detect Depth of

3.2 Tracking object with frames Coordinate Feature Positioning
system Points
A plane is a flat surface represented by pose, dimensions, and
boundary points. The boundary points are convex. The plane
manager creates models for each detected plane in the
environment which can be added, updated, and removed. For Match
each frame, the change of plane event may be invoked with Features
planes that have been modified (added, updated, and
removed) since the last frame.

The AR plane manager instantiates a prefab plane to Frame 1
represent the plane upon detection. The prefab asset type acts Detect Depth of
as a template from which new model instances can be created Feature Positioning
in the scene and it stores an interior model complete with its
corresponding components and properties. In the default
mode, the AR plane contains a mesh visualizer and its
corresponding components for the rendering. The base Figure 2 – Module for tracking object with frames

– 41 –