29-06-2012, 01:36 PM
3D Modeling Turntable
[attachment=26096]
INTRODUCTION
3D modeling is prevalent in several fields, ranging
from architecture to video game development. However,
most 3D modeling is done using complicated 3D
software such as 3D Studio MAX, Google Sketchup, or
SolidWorks. A solution that can accurately model reallife
3D objects would be of significant benefit to
architects who want to render their wooden/cardboard
models and game developers who want to add real-life
detail into their video game levels. Additionally, a 3D
modeler would prove useful to online retailers who wish
to provide customers with a preview of how an object
looks. A vendor can show how a pair of sunglasses look
from all sides and even overlay the 3D model on a
customer.
DESIGN STRATEGY AND IMPLEMENTATION
Team Supermodel's design strategy focused on taking
advantage of our team's multi-disciplinary nature and
used elements of Computational and Applied Math
(CAAM), Electrical and Computer Engineering (ECE),
and Mechanical Engineering (MECH). While our
CAAM team member championed the algorithmic
portion of the design, the ECE team member focused on
the electronics interface, and the MECH student worked
on the functional and aesthetic underlying mechanical
design.
Calibration
In order to gather the desired 3D volumetric data, basic
geometric information about each camera is required.
We used Tsai’s method of calibrating a camera with a
single-view set of coplanar points[1]. The method, in
brief, is to solve a system of equations that is set up
based on the observed image coordinates of a set of
points and the real world position of those points using
known equations for projections of 3D points to a 2D
plane.
Image Gathering
We decided to use a webcam to capture images so that
we could drive down cost and easily pull image data as
needed. Looking primarily for a low-cost webcam, we
decided on using the Philips SPC 230NC webcam.
Upon initial testing of the algorithms using one
webcam, we realized that one vantage point was not
enough to discern crucial geometric information about
an object. As such, we decided to use a secondary
webcam pointed at a different angle to provide additional
image information.
[attachment=26096]
INTRODUCTION
3D modeling is prevalent in several fields, ranging
from architecture to video game development. However,
most 3D modeling is done using complicated 3D
software such as 3D Studio MAX, Google Sketchup, or
SolidWorks. A solution that can accurately model reallife
3D objects would be of significant benefit to
architects who want to render their wooden/cardboard
models and game developers who want to add real-life
detail into their video game levels. Additionally, a 3D
modeler would prove useful to online retailers who wish
to provide customers with a preview of how an object
looks. A vendor can show how a pair of sunglasses look
from all sides and even overlay the 3D model on a
customer.
DESIGN STRATEGY AND IMPLEMENTATION
Team Supermodel's design strategy focused on taking
advantage of our team's multi-disciplinary nature and
used elements of Computational and Applied Math
(CAAM), Electrical and Computer Engineering (ECE),
and Mechanical Engineering (MECH). While our
CAAM team member championed the algorithmic
portion of the design, the ECE team member focused on
the electronics interface, and the MECH student worked
on the functional and aesthetic underlying mechanical
design.
Calibration
In order to gather the desired 3D volumetric data, basic
geometric information about each camera is required.
We used Tsai’s method of calibrating a camera with a
single-view set of coplanar points[1]. The method, in
brief, is to solve a system of equations that is set up
based on the observed image coordinates of a set of
points and the real world position of those points using
known equations for projections of 3D points to a 2D
plane.
Image Gathering
We decided to use a webcam to capture images so that
we could drive down cost and easily pull image data as
needed. Looking primarily for a low-cost webcam, we
decided on using the Philips SPC 230NC webcam.
Upon initial testing of the algorithms using one
webcam, we realized that one vantage point was not
enough to discern crucial geometric information about
an object. As such, we decided to use a secondary
webcam pointed at a different angle to provide additional
image information.