1
Digital Representation of Physical Artifacts: The Effect of Low Cost, High Accuracy 3D Scanning
Technologies on Engineering Education, Student Learning and Design Evaluation
DETC2013-12651
Nitish Vasudevan & Conrad S. Tucker {nuv115, ctucker4}@psu.edu
Tuesday, August 6th, 2013
Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/
2
OverviewIntroduction• Research Motivation
Methodology• Research Objective
• Proposed Methodology
• Data Acquisition
• Quantifying Form Similarity
Case Study
• Experiment Setup – 3D Scanner
• Survey Results
• Student Grades
ConclusionVasudevan, Tucker 2013 www.engr.psu.edu/datalab/Overview
Introduction
3Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Introduction
4
Research Motivation
• Educators prefer hands-on activities to convey concepts
easily to students*.
• Increasing use of technology creates lack in hands-on
activities to supplement digital student learning.
• With advancements in digital interactive technology, with
products such as Microsoft’s Kinect and Nintendo Wii, the
boundaries of hands-on interaction are being pushed into
the digital space.
Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Introduction
* Gorman et al.
Shift from physical to digital space
5
• Experiments conducted by Stanford University with their
bicycle disassembly course have shown that students
learn better when multimedia tools are employed as part
of the existing teaching methodology*.
• Digital tools such as the 3D scanner, livescribe digital pen
and the voice to text feature have accelerated and
improved the process of information extraction through
simple hands-on activities.
Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Introduction
* Regan and Sheppard
6Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Introduction
• Variations across digital model assessments given by
graders (teaching assistants). Increase in student designs,
increases the variations in grades and the number of
graders.
• Automated grading solutions have focused on textual and
choice based response data to a large extent*.
Techniques to quantify qualitative design data has not
been explored.
Research Motivation – contd. Automated student assessment techniques
* Zoeckler and Valenti et al.
Methodology
7Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Methodology
• To help bridge the gap between digital design education and
hands-on experiences in design classrooms.
• Assess variations in grading across student design activities.
8Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Methodology
Research Objective
9Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Methodology
3D Models
Student Designs
Scanner Models
Data Acquisition
Form Similarity
Student Grading
Design Software
3D ScannerPhysical
InteractionDigital
Interaction
Methodology Flowchart
10
Proposed Methodology
• Introduce 3D scanning as a technique to supplement
digital model creation through hands-on interaction
with live artifacts.
• Use degree of form similarity as a measure to evaluate
student designs benchmarked against a standard model.
Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Methodology
• Sample 3D Scanner.
Vs.
• Sample object to object comparison.
11
facet normal ni nj nkouter loopvertex v1x v1y v1z
vertex v2x v2y v2z
vertex v3x v3y v3z
endloopendfacet
facet normal ni nj nkouter loopvertex v1x v1y v1z
vertex v2x v2y v2z
vertex v3x v3y v3z
endloopendfacet
3D Models -Stereo lithography
files
Design 1
Design 2
Design 3
Wire-mesh
Data Acquisition
Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Methodology
12
Data Acquisition – contd.
• 3D scanners are digital image and depth capturing devices
which help create digital 3D models of objects.
• They consist of a combination of photographic lenses to
capture images and lasers to capture point by point depth
values across the surface of the artifact.
Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Methodology
Object to be scanned
3D Scanner
Digital model of object
13
3D Scanner - Sample
Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Methodology
• 3D Scanner in a classroom setup along with generated 3D models of objects.
14
Form Similarity• Generated student 3D models (STL files) are compared
with the model created by the scanner using the form
similarity metric using the evaluation of reeb graphs as
generated by Doraiswamy et al.
• Form similarity by definition is the evaluation of degree
of alikeness in form (pure geometric) between two
artifacts.
• The result gives a relative value of the deviation from
standard when comparing various models to a
benchmarked model.Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Methodology
15
Reeb Graphs• Reeb graph is a form visualization technique based on
Morse theory which evaluates geometry of objects
based upon the surface topology through the
determination of iso-surface parameters along
increasing level set values (along Z-axis).
• Similarities in Reeb graphs
represent similarities in
models through the
analysis of identical critical
points in the Reeb graph.*Doraiswamy et al.
Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Methodology
16
• Reeb graphs can be compared for similarities through a
comparison of the level set data for each reeb graph.
• The degree of similarity is a direct correlation to the
level of similarity between the two 3D models.
Level set data
Saddle Maxima Minima
1 0 0
2 0 2
3 6 5
. . .
1543 1554 1023
Reeb Graphs – contd.
Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Methodology
• Sample of generated data.
• Reeb graph comparison –visualization.
17
Quantitative Assessment
• Student designs created are compared for similarities in
form using the reeb graph technique.
• Comparisons are done between student designs and a
benchmark design either generated from a 3D scanner
or present in the product database.
• Student grades are a direct representation of the
degree of similarity between the comparisons (e.g.
similarity value of 0.86 correlates to a grade of 86/100).
Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Methodology
18
Case Study
Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Case Study
19
Object
Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Case Study
Data Set
• Models generated by Students on Solidworks.
• Model generated by the Nextengine® 3D Scanner.
20Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Case Study
Design Assessment
• 25 Students created the model of thecoffee mug during a lab class of 2 hoursduration.
• Each student was given identical mugs andwere asked to recreate the model usingthe standard Solidworks design package.
21
On a scale of 5 (5 being the easiest):
Q1: How much easier was it to regenerate a solid part using the
scanner as opposed to designing it on the software?
Q2: How easy was it to learn the working of the scanner?
Q3: How easy was it to navigate through the software that is
associated with the scanner for scanning?
Fre
qu
en
cy
54321
10.0
7.5
5.0
2.5
0.0432
20
15
10
5
0
432
20
15
10
5
0
Question 1 Question 2
Question 3
Question 1
3.28
StDev 0.5416
N 25
Question 3
Mean 3.16
StDev 0.5538
Mean
N 25
3.44
StDev 1.003
N 25
Question 2
Mean
Histogram of Question 1, Question 2, Question 3Normal
• Additional Survey Questions-3D Scanner
Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Case Study
22
Student Student Model TA 1 Score TA 2 Score Scores from reeb graph
1
87 69 91
2
78 47 78
3
95 90 100
4
92 73 49
5
95 84 76
6
90 68 74
7
80 68 98
8
92 78 99
Results for student assessment
Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Case Study
23
Statistic TA 1 TA 2
Mean 88.63 72.13
Std. Deviation 6.5 12.93
One sample t at 95% CI (83.19,94.06) (61.31,82.94)
Estimate of difference 16.5
95% CI for difference (5.0968, 27.9032)
T – value 3.22
P – value 0.009
Fre
qu
en
cy
1201101009080706050
4.8
3.6
2.4
1.2
0.0
1201101009080706050
4.8
3.6
2.4
1.2
0.0
TA 1 Reeb Graph
TA 2
TA 1
83.13
StDev 17.50
N 8
TA 2
Mean 72.13
StDev 12.93
Mean
N 8
88.63
StDev 6.501
N 8
Reeb Graph
Mean
Histogram of TA 1, Reeb Graph, TA 2Normal
• Results from Student design assessment
Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Case Study
24Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Conclusion
Conclusion
2525Vasudevan, Tucker 2013 www.engr.psu.edu/datalab/Conclusion
• The conducted research aims to aid in bridging the gap
between digital and hands-on activities in design
classroom by introducing 3D scanners as a tool to create
digital models of objects.
• Variations in assessments of student generated models
across graders is eliminated through the automated
evaluation of models by similarity comparison to a
benchmarked model.
26
QuestionsComments