World Wide Web by using a simple php code. The web page is http://3dsite.dhs.org/~dynamic. The user can upload a 3D model and the system will calculate.
A 3D Object Retrieval System Based on Multi-Resolution Reeb Graph Ding-Yun Chen and Ming Ouhyoung Department of Computer Science and Information Engineering, National Taiwan University, Taipei, Taiwan E-mail: {dynamic, ming}@cmlab.csie.ntu.edu.tw Abstract
models retrieval using the distribution of moment,
This paper proposes a 3D model retrieval
normal, cord, color, material and texture. Zhang and
system which extended the work of Hilaga in 2001.
Chen [2] propose a 3D model retrieval system using
We apply the pre-processing stage to 3D models in
volume-surface ratio, aspect ratio, moment invariants
practical use. The demo system with over 450 3D
and Fourier transformation coefficients. Elad et al. [6]
models from the Net is on the web page:
apply relevant feedback to 3D object retrieval, which
http://3dsite.dhs.org/~dynamic, and can also be used
uses moments as features. Osada et al. [4] propose and
in PocketPC with wireless LAN card. There are 445
analyze a method for computing shape signatures for
various models in our database.
arbitrary 3D polygonal models. Hilaga et al. [1] propose a technique in which similarity between
1.
Introduction In this decade, multimedia data, which usually
polyhedral models is quickly, accurately, and automatically calculated by comparing the skeletal and
doesn’t need any text to represent, grow rapidly. Content-based retrieval for multimedia data becomes
topological structure. The structure decomposes 3D model to a one-dimensional graph structure. The graph
more and more important. In order to communicate with people’s information, the MPEG group aims to
is invariant to translation, rotation and scaling, robust against connectivity changes, and resistant against
create MPEG-7 international standard, also known as “Multimedia Content Description Interface”, for the
noise, certain changes due to deformation. Our system is based on the research of Hilaga,
description of the multimedia data, including image, video, audio, 2D shape and 3D object [8]. 3D object
which is one of the best ideas among the previous works. Fig. 1 shows the flow chart of our system
retrieval research is active now, because the technique of 3D modeling and digitizing tools is on a
when querying by a 3D object. The last two stages are the same with Hilaga, please refer to [1]. Chapter
progressive improvement. In the last few years, several articles have been devoted to the study of 3D object retrieval. Cyr and
2 details the pre-processing stage in our system.
2.
Pre-Processing of 3D Object
Kimia [5] proposed an aspect-graph approach. They
This stage is designed for accurately and fast
generate a set of 2D silhouette for each 3D object, and
getting the search key. There are four steps in this
then measure the similarity between two views by 2D
stage: merging vertices, merging parts, re-sampling
shape similarity metrics. Kolonias et al. [3] proposed
and adding short-cut edges. The first two steps solve
aspect ratio, a binary 3D shape mask and set of paths
practical problems when many models are used. The
outlining the shape of the 3D object for matching.
last two steps are modified from the approach of
Paquet and Rioux [7] presented an approach for 3D
Hilaga [1].
averagely split into many segments. Re-sample by
Input a 3D model
connecting the new vertices from the two edges. In Merge vertices Merge different parts Re-sample 3D model
Stage 1: Pre-processing is used for accurately and fast calculating the search key (Chapter 2)
Add short-cut edges Geodesic distance Create MRG Compare models Show results
the third case, if all three edges in triangle are larger than threshold, select the two largest edges and split into many triangles first, then each triangle recursively is split using the second case. The fourth step is adding some edges called
Stage 2: Calculate the search key MRG for the 3D model Stage 3: Compare the MRG to all models in the database
Fig. 1 The flow chart of our system. The same vertices may be shared in duplicate at
"short-cut edges" to models. The purpose of short-cut edges is used for accurately calculating the search key. The short-cut edges make the distance of two vertices in adjacent triangles to be straightforward through the 3D surface, as shown in Fig. 2 (a). There are two cases of triangles. In the first case, if a triangle have two same vertices with other triangle. Take Fig. 2 (b) as an example, triangle 1 and 2 share ace, which
adjacent triangles for 3D models. The first step is to
vertex c and d, and then get the angle
merge these vertices. The way we used is to sort the
is sum of
vertices according to the coordinate, and then check
180 degree, add a short-cut edge between vertex a
the adjacent vertices. The time complexity is O(nlgn),
and e. Alternative can start from the angle
since we use the heapsort algorithm.
The angle is less than 180 degree means that edge ae
acd and
ecd. If the angle is less than
ade.
3D models are saved as different parts in many
will inside the polygon aced. The distance of
cases for easy editing and animation. The approach
short-cut edge ae can be calculated by the following
we used is to construct the search key by the whole
formula. Given the length of two edges (x, y) and
model rather than some separated parts. Therefore,
their angle (w), then the length of third edge is:
we add edges to connect different parts. The
x 2 + y 2 − 2 ⋅ x ⋅ y ⋅ cos( w) .
approach is similar to the first step. The coordinate of
to connect the same quantized value for the separated parts. The time complexity is also O(nlgn).
b
a
each vertex is quantized first, and an edge is created
3 2
c
1
d
For more accurate calculation of the search key, each triangle has to be re-sampled. Each triangle splits into smaller triangles until all edges are less than a threshold. To speedup re-sampling time in all
e (a) (b) Fig. 2 The short-cut edges make the distance of two vertices in adjacent triangles to be straightforward through the 3D surface.
cases, each triangle is split at once. There are three cases for all triangles. In the first case, there is only one edge in triangle larger than threshold, just split into two triangles. In the second case, if two edges in a triangle are larger than threshold, the two edges are
In the second case, if some triangle have only one the same vertex with this triangle, and if there is a triangle adjacent to the two triangles. For example, triangle 1 and 3 share vertex d, and triangle 2 adjacent to them. Then get the angle
edb, which is
sum of
edc,
cda and
adb. As similar to case
[1]
M. Hilaga, Y. Shinagawa, T. Kohmura and T. L.
one, if the angle is less than 180 degree, add a
Kunii, “Topology Matching for Fully Automatic
short-cut edge between vertex b and e.
Similarity Estimation of 3D Shapes”, ACM SIGGRAPH, pp. 203-212, Aug. 2001.
3.
Experimental Results
[2]
The system is implemented in C language, and
C. Zhang and T. Chen, “Efficient Feature Extraction for 2D/3D Objects in Mesh
is compiled using gcc in Linux. The system is on the Representation”, IEEE International World Wide Web by using a simple php code. The
Conference on Image Processing, Oct. 2001.
web page is http://3dsite.dhs.org/~dynamic. The user [3] can upload a 3D model and the system will calculate
I. Kolonias, D. Tzovaras, S. Malassiotis and M. G. Strintzis, “Fast Content-Based Search of
the MRG for the model, and then compare it with all VRML Models Based on Shape Descriptors”, models in the database. Alternatively, the user can
IEEE International Conference on Image
select one model in the database. The system Processing, Oct. 2001. compares it with all other models and the results
[4]
R. Osada, T. Funkhouser, B. Chazelle and D.
show thumb pictures of top similar models. There are Dobkin “Matching 3D Models with Shape 445 various models, downloaded from [10] and [11],
Distributions”, Workshop on Shape-Based
in the database. The average time of comparing two Retrieval and Analysis of 3D Models, Oct. 2001. models is about 0.08 second in a PC with Pentium III
[5]
C. M. Cyr and B. B. Kimia, “3D Object
800MHz CPU. Fig. 3 shows results of 3D object Recognition Using Shape Similiarity-Based retrieval in our system. In addition, our retrieval
Aspect Graph”, International Conference on
system can be used in PocketPC with wireless LAN Computer Vision, 2001. card, as shown in Fig. 4.
4.
[6]
Retrieval of VRML Objects – A Iterative and
Discussion and Future Works While the system can work well in many cases,
there exist considerable improvements as to the following problems:
Interactive Approach”, 2001. [7]
sub-graph matching, partial
E. Paquet and M. Rioux, “Content-Based Access of VRML Libraries”, Lecture Notes in
matching, MRG doesn't always represent the skeletal structure, different density of vertices, some vertices
M. Elad, A. Tal and S. Ar, “Content Based
Computer Sciences, Vol.1464, pp. 20-32, 1998. [8]
S. Jeannin, L. Cieplinski, J. R. Ohm and M.
inside the 3D object. To overcome the problems, we
Kim, MPEG-7 Visual part of eXperimentation
plan to use the hierarchical medial axis to match the 3D
Model Version 7.0, ISO/IECJTC1/SC29/WG11
models. The 3D medial axis algorithm based on radial
/N3521, July 2000.
basis function (RBF) is proposed in our group [9].
[9]
F.C. Wu, W.C. Ma and M. Ouhyoung, “Skeleton Extraction of 3D Objects with
5.
Conclusion
Radial Basis Function”, Technical Report
A 3D object retrieval system, which improves NTUCSIE 02-01, Dept. of CSIE, National the practical use of Hilaga’s research [1], is proposed.
Taiwan University, Taipei, Taiwan, Apr. 2002.
The pre-processing stage is applied for accurately [10] http://www.3dcafe.com and fast getting the search key.
Reference
[11] http://www.3dm-mc.com
Fig. 3 Some results of 3D object retrieval in our system.
(continue)
Fig. 4 Our retrieval system can be used in PocketPC with wireless LAN card.