DRR2014 StructureAnalysisforPlaneGeometryFigures

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/266203349

Structure Analysis for Plane Geometry Figures

Conference Paper in Proceedings of SPIE - The International Society for Optical Engineering · February 2014
DOI: 10.1117/12.2042462

CITATIONS READS
6 3,504

5 authors, including:

Xiaoqing Lyu lu Liu


Peking University National University of Singapore
70 PUBLICATIONS 381 CITATIONS 18 PUBLICATIONS 119 CITATIONS

SEE PROFILE SEE PROFILE

Zhi Tang
Sichuan University
192 PUBLICATIONS 2,897 CITATIONS

SEE PROFILE

All content following this page was uploaded by lu Liu on 30 September 2014.

The user has requested enhancement of the downloaded file.


Structure Analysis for Plane Geometry Figures

Tianxiao Feng1,a, Xiaoqing Lu2,a,b, Lu Liua, Keqiang Lia , Zhi Tanga,b


a
Institute of Computer Science & Technology, Peking University, Beijing, China; bState Key
Laboratory of Digital Publishing Technology (Peking University Founder Group Co., Ltd), Beijing,
China

ABSTRACT
As there are increasing numbers of digital documents for education purpose, we realize that there is not a retrieval
application for mathematic plane geometry images. In this paper, we propose a method for retrieving plane geometry figures
(PGFs), which often appear in geometry books and digital documents. First, detecting algorithms are applied to detect
common basic geometry shapes from a PGF image. Based on all basic shapes, we analyze the structural relationships between
two basic shapes and combine some of them to a compound shape to build the PGF descriptor. Afterwards, we apply matching
function to retrieve candidate PGF images with ranking. The great contribution of the paper is that we propose a structure
analysis method to better describe the spatial relationships in such image composed of many overlapped shapes. Experimental
results demonstrate that our analysis method and shape descriptor can obtain good retrieval results with relatively high
effectiveness and efficiency.

Keywords—plane geometry figure, structure analysis, compound shape, document image retrieval

1 INTRODUCTION
Computer-aided instruction has become increasingly popular in recent years. Consequently, a growing number of
teaching contents have been digitalized and stored electronically. Thus, the subject of pattern recognition for document
images or figures has become important. Plane geometry figure (PGF, Figure 1) is a type of document graph. To the best
of our knowledge, existing plane geometry question retrieval systems generally focus on the keyword in the text of a
plane geometry problem. However, the description of such keywords may not represent the questions sufficiently, thus
leading to inaccurate performance.

Figure 1 Examples of Plane Geometry Figures.

One key issue of graph recognition is to explore exquisite descriptors that reflect the nature of a shape. Description
techniques can be generally classified into two classes: region-based methods and contour-based methods [1]. Region-
based features are primarily intended for overall statistics or the analysis of the entire region of a shape, such as moment
invariants [2] or transforms [3, 4]. Although these methods are not susceptible to noise, which lead to satisfactory
discrimination results, they fail to capture the structural features of a shape. By contrast, contour-based methods [5, 6]
1
paddy5625@gmail.com
2
lvxiaoqing@pku.edu.cn

Document Recognition and Retrieval XXI, edited by Bertrand Coüasnon, Eric K. Ringger,
Proc. of SPIE-IS&T Electronic Imaging, Vol. 9021, 90210R · © 2014 SPIE-IS&T
CCC code: 0277-786X/14/$18 · doi: 10.1117/12.2042462

SPIE-IS&T/ Vol. 9021 90210R-1


perform better in single contour shapes; however, these techniques can hardly describe complex inner structure and
geometric attributes.
Decomposing PGFs requires effective detection of shapes. Ballard [7] proposed a general detection method based on
Hough transform. However, this method involves a highly complex computation process. Afterwards, numerous special-
purpose detectors have been invented, such as circle and line-segment detectors, which are required in PGF recognition.
Other circles detectors [9–11] are used for natural images and document graphs. Lamiroy [8] proposed a quick and
robust circle detection method especially developed for document graphs. Duda and Hart [12] were the first to propose a
line detection method. Lin and Nevatia [13] invented a method which can effectively splice a rectangle by using detected
line segments. Nayef [14] presented a complete symbol retrieval system based on a collection of line drawings. Recently,
Li and Lu[15] proposed a method to detect overlapped quadrangles in PGFs.
Based on the above methods, we first apply detection algorithms to extract basic geometric shapes. After that, in this
paper, we focus on the structure analysis for PGF images. To illustrate the features of a PGF, our proposed descriptor
mainly focuses on structural relationship analysis of geometric elements after they were detected. We analyze the
relationships between two basic shapes and describe the structural features in PGFs with compound shape elements. The
remainder of this paper is organized as follows. Sec. 2 introduces the proposed descriptor for PGFs. Sec. 3 establishes
the match rule for PGF retrieval. Sec. 4 presents the experiments conducted in the study and provides a summary of the
research.

2 THE DESCRIPTOR FOR PGF


2.1 Basic Shape Detection
PGFs are different from other shape images for they are composed of basic shapes, such as circles, triangles, and
quadrilaterals, as shown in Figure 1. Given a PGF, our aim is to decompose it into several basic shapes. Therefore, we
need to detect all basic shapes in the figure. To represent the detected result, a set G = {S1 ,..., S n } is used, wherein
element S is a basic shape, which can be a circle, a rectangle, a parallelogram, a trapezoid, or a line.

Figure 2 An example for basic shape detection: one circle and three triangles have been detected in given plane geometry figure.

Given an example showed in Figure 2, we can see one circle and three overlapped triangles have been detected.
The main detection steps are listed in Table 1.

Table 1 Detection Steps for Figure G.


1. Circles are detected [9] and added to set G.
2. Lines are detected through Hough transform and line segments are found.
3. Obtained line segments are used to splice rectangles, parallelograms, and trapezoids [15].
4. Discovered line segments are used to splice triangles.
5. Shapes found in steps 3 and 4 are added to set G.
6. Unused line segments are added to set G.
In some cases PGFs cannot be discriminated by counting and comparing their basic shapes only. As an example,
each of the three PGFs in Figure 3(a) contains two circles and one triangle but they look quite different. In contrast,
those PGFs in Figure 3(b) look similar with almost the same arrangement of elements but different quantities of triangles.
Therefore, the relationships between basic shapes in one PGF should not be ignored, and it is essential to find the
structural features in PGFs.

SPIE-IS&T/ Vol. 9021 90210R-2


(a) (b)
Figure 3 (a)Three different PGFs with same quantities of basic shapes; (b)Three similar PGFs

2.2 Relationship between Two Basic Shapes


It is necessary to analyze the relationships between basic shapes before finding the structural features in PGFs. Let S1
and S2 be the two basic shapes in PGF G. Afterwards, we will discuss the relationship of various types of S1 and S2.

2.2.1 Triangles and Quadrilaterals


If S1 and S2 are both triangles, then let a1, a2, and a3 be the three edges of S1; and b1, b2, and b3 be the edges of S2. We
can calculate a 3×3 matrix to represent the relationships between two triangles.

 Rss ( a1 , b1 ) Rss ( a1 , b2 ) Rss ( a1 , b3 ) 


 
M 33 =  Rss ( a2 , b1 ) Rss ( a2 , b2 ) Rss ( a2 , b3 ) 
 Rss ( a3 , b1 ) Rss ( a3 , b2 ) Rss ( a3 , b3 ) 
(1)
where Rss ( ai , b j ) ( i = 1, 2,3; j = 1, 2,3) represents the relationships between two segments, and the values are defined
in Table 2.

Table 2 Values of the Relationships between Two Segments.

Relationships Examples Rss ( ai , b j )


Absolutely congruent Omitted 10
P is not in the middle 9
One common endpoint
P is in the middle 8
Collinear
Included 7
No common endpoint Partially Overlapped 6
Separated 5
Parallel 4
L-junction 3
Intersected T-junction 2
Not Parallel
X-junction 1
Not intersected 0

We also use a 4×4 matrix to represent the relationships between two quadrilaterals; and a 4×3 matrix to represent
the relationships between a quadrilateral and a triangle. Hereafter, a quadrilateral can only be a rectangle, a trapezoid, or
a parallelogram.
M 44 ( i, j ) = Rss ( ai , b j ) i = 1...4, j = 1...4
(2)
M 43 ( i, j ) = Rss (ai , b j ) i = 1...4, j = 1...3 (3)

SPIE-IS&T/ Vol. 9021 90210R-3


In Equations (2) and (3), M pq ( i, j ) represents the ith row and jth column in the matrix Mpq (p = 3,4; q = 3,4; and p
>= q).
We use Mpq to reflect the relationships between each edge of S1 and each edge of S2, where S1 and S2 can be a triangle
or a 4-edge shape. However, the matrix contains every possible mapping situation, Thus, in order to normalize the
descriptor Mpq, we find the best mapping vector R to describe the relationship between the two shapes. The algorithm is
shown in Table 3.

Table 3 Algorithm 1: find the best mapping vector R from Mpq


for k = 1:q
Find the max number Ak in Mpq.
Remove the line and column of Mpq where Ak located.
End
Thus, we obtain the relationship between two shapes as:
R ( S1 , S 2 ) = [ A1 ... Ak ]
(4)

000
M= 3 3 4

3 8 0

Figure 4 The matrix which represents the relationship between two given triangles

For example, as shown in Figure 4, the matrix M is the origin representation of given two triangles in the left. The
number 8 are the max number in the matrix which represents the relationship between the segment a1 and b1. After we
removed the line and column of the matrix where the number 8 located, the rest max number is 4, which represents the
parallel segments a2 and b2. Since a3 and b3 are neither parallel nor intersected, their relationship score is zero. The
relationship between these two triangles are shown in Equation (5).

(5)
If S1 and S2 are both quadrilaterals, then R will have four values; otherwise, R will only have three values. We use this
method to represent the relationships between triangles or quadrilaterals. However, this method is not suitable for
representing the relationships between a circle and another basic shape.

2.2.2 Relationships between a Circle and a Quadrilateral


Let S1 be a circle and S2 be a quadrilateral in PGF G. We count the number of vertices for S2, which are found along
the circle S1 (denoted as Nv), and the number of tangential points on S2 to the circle S1 (denoted as Ne), as follows:
N v = card { p | p ∈ S 2 .vertex ∧ dist ( p, S1.c) = S1.r}
(6)
N e = card {e | e ∈ S 2 .edge ∧ dist (e, S1 .c) = S1 .r}
(7)
In Equations (6) and (7), S1.c denotes the centroid of S1, whereas S1.r defines the radius. The relationship between the
circle and the quadrilateral is defined in Equation (8):
R ( S1 , S2 ) = ( N v , N e ) (8)
2.2.3 Relationships between a Circle and a Triangle
Let S1 be a circle and S2 be a triangle in PGF G. Aside from Nv and Ne, which have been defined in Equations (6) and
(7), we introduce two extra properties, Nc and Nd , as follows:

SPIE-IS&T/ Vol. 9021 90210R-4


1 ∃p ( p ∈ S2 .vortex ∧ dist ( p, S1.c) = 0 )
Nc = 
0 others (9)

1 ∃e ( e ∈ S2 .edge ∧ e is a diameter of S1 )
Nd =  (10)
0 others

The relationship between the circle and the triangle is defined in the following equation:
R ( S1 , S 2 ) = ( N v , N e , N c , N d ) (11)
2.2.4 Relationships between Two Circles
Let d represent the distance between the two centroids of the circles, and r1 and r2 represent their radii. Table 4 shows
five scenarios in which we compare d, r1+ r2, and |r1- r2|.

Table 4 Relationships between Two Circles.


Condition Circumstance
d > r1 + r2 Separated
d = r1 + r2 Circumscribed

r1 − r2 < d < r1 + r2 Intersected

d = r1 − r2 Inscribed

d < r1 − r2 Included
Theoretically, we could analyze all relationships between any two basic shapes with the method introduced in Sec.
2.2, as the number of types of basic shapes is finite. However, the computation will be huge if we enumerate all
relationships. Moreover, not all kinds of relationships are equally important for the structural feature description of
PGFs. It is necessary to select those important relationships and exclude unimportant ones. In the next section,
compound shapes will be adopted to fulfill this idea.

2.3 Compound Shapes


Discussing all kinds of relationships between two basic shapes will encompass a large and complex project. However,
most relationships are weak, thus, finding strong relationships among these shapes is more important than enumerating
all of them. In this paper, Through the analysis and a set of experiments, we finally find out some types of compound
shape with strong relationships that are important to identify different PGRs. In the following, we analyze the
relationships between two circles, a circle and a triangle, and a circle and a quadrilateral, as representative relationships
to elaborate our methods.

2.3.1 Two Circles


As generally defined, there are five kinds of relationships between two circles: “circumscribed,” “intersected,”
“inscribed,” “included,” and “separated” (as defined in Table 4, and all of which have been defined in Sec. 2.2.4).
However, here we only consider the first four “strong” relationships and discard the “separated” relationship as “weak”
relationship, which contributes very little to the structural features.

2.3.2 A Circle and a Triangle


Based on the experiments, six kinds of strong relationships between a circle and a triangle show their significance
compared to other weak relationships, as shown in Table 5. Generally speaking, the six strong relationship have one
point in common is that the triangle must at least share one vertex with its circles, which increases the importance of the
compound shape.

SPIE-IS&T/ Vol. 9021 90210R-5


Table 5 Compound shapes-A Circle and A Triangle.

Nv 3 Nv 3 Nv 2
Ne 0 Ne 0 Ne 0
Nc 0 Nc 0 Nc 1
Nd 1 Nd 0 Nd 0
Nv 2 Nv 1 Nv 0
Ne 1 Ne 1 Ne 3
Nc 0 Nc 1 Nc 0
Nd 1 Nd 0 Nd 0
2.3.3 A Circle and a Quadrilateral
Quadrilaterals can be a rectangle, a trapezoid, or a parallelogram. In the case of a rectangle, we use the condition
N v + N e >= 4 , which can generate four situations. For a trapezoid, we only choose a condition when all the vertices of
the trapezoid fall within the circumference of the circle. We discard all the relationships between a circle and a
parallelogram.
Through analysis and a set of experiments of combinations of two shapes, 41 compound shapes are looked on as
strong relationships, as shown in Figure 5. Generally, both compound and basic shapes are used to represent the feature
of a PGF.
Consequently, we obtain the updated set G = { P1 ,..., Pn } ; in this set, P can be a basic or compound shape. We
maintain the perimeter and the area in each P. If P is a compound shape, then we use the sum of the perimeters and areas

AAAAXA
s=== 0 p/
A
of the two basic shapes to represent those of the compound shape P.

AAAA AAA7 7
q,Th H
,

6
01810G C r a _J a00@c)i00i)
Figure 5 List of all the compound shapes.

Table 6 Algorithm 2: the algorithm which composes the compound shapes based on the basic shapes.
Sort the basic shapes (descending) by area.
x = 0;
for i = 1 to n
for j = 1 to i-1
if (Si is ‘used’ and Sj is ‘used’) continue; end (if)
if R ( Si , S j ) is strong (Figure 5)
x = x+1;
Generate a compound shape Dx;
Mark Si and Sj ‘used’;
end(if)
end(for)
end(for)
Remove ‘used’ basic shapes in set G.
Add compound shapes D1…Dx to set G.
Sort all the elements (descending) by area in set G.

SPIE-IS&T/ Vol. 9021 90210R-6


3 MATCH RULES
In this section, we introduce a function Match ( G1 , G2 ) , which has a higher value when G1 and G2 are more similar to
each other, and a lower value when G1 and G2 are not. Let G1 { P1 ,..., Pm } and G2 {Q1 ,..., Qn } be two figures. The first step
is to define a weight between an element P in G1 and an element Q in G2. Given that each element can either be a basic
or compound shape, categorized discussions are proposed below.

3.1 Weight between Two Basic Shapes


We evaluate the significance of different kinds of basic shapes based on their appearance rate. We first count the
total numbers of basic shapes in our datasets and evaluate them manually as visual saliency or not. Then we calculate
the rates of visual salient shapes, as shown in Table 7. The rates reflect the visual importance for different types of basic
shapes. In Table 7, we can see all the circles in PGF are visual salient, which means circles in PGF are more significant
than other triangles and quadrangles. On the other hand, triangles in PGF are less visual salient. The datasets are
introduced in Sec. 4.

Table 7 the quantity of visual salient basic shapes and total basic shapes in our dataset
Circles Triangles Rectangles Trapezoids Parallelograms
Visual salient quantity 219 134 31 26 5
Total quantity 219 1955 38 65 40
Rate 100% 6.9% 82% 40% 13%
Let P be a basic shape in figure G1 and Q be a basic shape in figure G2. We assign the class weights as 50 times of
the rates in Table 7 under the condition that P and Q are same types of basic shapes, which reveals their different
significance. In our actual experiments, the weights C ( P, Q ) are replaced with close integers as shown in the last
column in Table 8.
On the other hand, if P and Q are different types, we assume that they are not matched so that the weight C ( P, Q )
is set to zero.

Table 8 Class weights for basic shapes P and Q

P Q Rates times 50 C ( P, Q )
Circle Circle 50 50
Rectangle Rectangle 41 40
Trapezoid Trapezoid 20 20
Parallelogram Parallelogram 6.5 7
Triangle Triangle 3.45 3
Different types -- 0
We also consider the partial circle rate of P and Q as
A ( P, Q ) = 1 − ( PCR ( P ) − PCR ( Q ) ) * 2 π
(12)
where
Area ( S )
PCR ( S ) = , S ∈ { P, Q}
Perimeter ( S ) (13)

We should also generate a size factor to keep the scale invariance, as follows:
Area ( P ) / Area ( Q ) 1
S ( P, Q ) = if S ( P, Q ) > 1, S ( P, Q ) = (14)
Scale ( P ) / Scale ( Q ) S ( P, Q )

Equation (14) indicates that S ( P ) is the biggest basic shape S1 of the set G1 which can reflect size. The values of
both A and S are obviously between 0 and 1. Subsequently, we define the weight between P and Q as

SPIE-IS&T/ Vol. 9021 90210R-7


W ( P , Q ) = C ( P, Q ) * A ( P , Q ) * S ( P, Q )
(15)
3.2 Weight between Two Compound Shapes
Let D1 be a compound shape in G1 and D2 be a compound shape in G2. We also use Equation (15) to calculate the
weight of D1 and D2, but redefine class match indices as

C ( D1 , D2 ) = Cd ( D1 , D2 ) + Ce ( D1 , D2 ) * max ( C ( S1 , S 2 ) + C (T1 , T2 ) , C ( S1 , T2 ) + C ( S 2 , T1 ) ) (16)


In Equation (16), Si and Ti are the two basic shapes which compose Di, and the parameter Cd ( D1 , D2 ) , Ce ( D1 , D2 )
are defined as:
N D1 and D2 are the same type
Cd ( D1 , D2 ) =  (17)
0 D1 and D2 are different types

2 D1 and D2 are the same type


Ce ( D1 , D2 ) = 
1 D1 and D2 are different types
(18)
We can rewrite Equation (16) as Equation (19). In Equation (19), if D1 and D2 are different kinds of compound
shapes, we use the maximum of class weights between the four basic shapes of two compound shapes. Otherwise, the
weight of D1 and D2 is much higher. According to the experiment results in Sec. 4.2, we set the parameter N as 40.

 N + 2* max ( C ( S1 , S 2 ) + C (T1 , T2 ) , C ( S1 , T2 ) + C ( S 2 , T1 ) ) D1 and D2 are the same type


C ( D1 , D2 ) = 
 max ( C ( S1 , S2 ) + C (T1 , T2 ) , C ( S1 , T2 ) + C ( S2 , T1 ) ) D1 and D2 are different types (19)

We also need to redefine the partial circle rate as

Area ( D ) Area ( S ) + Area (T )


PCR ( D ) = =
Circum ( D ) Circum ( S ) + Circum (T )
(20)
3.3 Weight between a Compound Shape and a Basic Shape
We need to redefine class match indices as shown in Equation (21) and use Equation (15) to calculate the weight as
follows:
C ( P, D ) = max ( C ( P, S ) , C ( P, T ) )
(21)
In Equation (21), P is a basic shape in G1, whereas D is a compound shape in G2, which are composed of S and T.

3.4 Match
We establish a match matrix for G1 { P1 ,..., Pm } and G2 {Q1 ,..., Qn } as follows:

 W ( P1 , Q1 ) W ( P1 , Q2 ) ... ... W ( P1 , Qn ) 
 
W ( P2 , Q1 ) W ( P2 , Q2 ) ... ... W ( P2 , Qn ) 
M ( G1 , G2 ) =  ... ... ... ... ... 
 
 ... ... ... ... ... 
W ( P , Q ) W ( P , Q ) ... ... W ( Pm , Qn ) 
 m 1 m 2
(22)

SPIE-IS&T/ Vol. 9021 90210R-8


We assume that Equation (22) is the weight matrix of a complete bipartite figure G = ( G1 , G2 ) = ( P1 ,..., Pm , Q1 ,..., Qn ) ,
and we find a maximum weight that perfectly matches G through the Kuhn-Munkres algorithm[16]. We use this
maximum weight as the of two PGFs G1 and G2, as follows:

KM ( G1 , G2 ) = Kuhn _ Munkras ( M ( G1 , G2 ) )
(23)
And finally obtain the matching score as:
2 * KM ( G1 , G2 )
Match ( G1 , G2 ) =
KM ( G1 , G1 ) + KM ( G2 , G2 ) (24)

4 EXPERIMENTS AND CONCLUSIONS


In this section, we demonstrate the effect of the description based on compound shapes and the match rules. For our
experiment, we use a computer consists of Intel Core i5 (3.20GHz), 4.00G RAM and Windows 7 operating system.

4.1 The datasets and the measure criterion


We have collected 267 plane geometry figures from five primary geometry books to build the image database. Each
image has a ground-truth set G = {S1 ,..., S n } to describe basic shapes set; thus, we can use the ground-truth data for the
detection evaluation. For the evaluation of retrieval, we conduct an online questionnaire between every two PGR images
for human to give a relevant score based on visual similarity.
We have set up a 267*267 matrix to score the similarity between each two PGFs. The similarity between each two
PGFs are voted manually with a score range of 0-5. We give 5 points for those very similar pair of PGFs while give 0
points for non-similar pair of PGFs. It can be concluded that the vote result matrix are a symmetric matrix, and the data
in the main diagonal of the matrix are meaningless since it just represent the similarity between a PGF and itself. Thus,
we just have voted 267*(267-1)/2=35511 pair of PGFs.

4.2 Experiment for the parameter N in Equation (19)


In our experiment, we use one plane geometry figure as the query, and process it with detection steps in Sec. 2.1 and
shape features in the rest of Sec. 2. Then we match it with another 266 PGFs in the datasets and find 5 PGFs with the
highest match indices. Then we sum the vote scores (declared in Sec. 4.1) with the query and the 5 found PGFs as the
experiment result.
Afterwards we use another plane geometry figure as another query and do the same procedure above until all 267
PGFs are queried. We use the sum of all vote scores to evaluate the effects of our method.
2330
2319
2316
2320 2312
2306
2310

2300 2295 295

2290

2280

2270
20 25 30 35 40 45 50

Figure 6 The experiment result for different parameter N in Equation (19)

The parameter N in Equation (19) (see Sec. 3.4) are set to 40 after we have tried the value of N=20,25,30,35,40,45,50.
The sum of all vote scores are shown in Figure 6.

SPIE-IS&T/ Vol. 9021 90210R-9


4.3 Experiment for description based on compound shapes
To verify the effects of the description, we use the Zernike Moments Descriptor (ZMD) method[17] for comparison
to evaluate the effects of the matching rules. We also use the original basic shape set (BSS) G = {S1 ,..., S n } , which
means the BSS method is similar to our method but skip the procedures described in Sec. 2.2, 2.3, 3.2, and 3.3. The
results show that our matching rules are effective because both the proposed method and the BSS perform better than the
ZM method.

ZMD

BSS
o e
OkeiCD
Proposed
method
00
Figure 7 Experimental results for the three methods

In Figure 7, the leftmost graph in each line is the query. The comparison of the second and the third graphs in each
line reveals that the proposed method and the BSS perform better than the ZMD method. We can also conclude from the
fourth and the fifth columns in Figure 7 that the proposed description based on the compound shapes perform better than
BSS because both the numerous triangles in the circle and their relationships are closer to the query.
The sum score after we use BSS and ZMD method are shown in Figure 8, we can see both our method and BSS
perform much better than ZMD method. The proposed method performs better than BSS because the BSS method lose
the relationships between basic shapes.

Figure 8 the sum score of Zernike, BBS and proposed method

We also sum the vote score of the queries and its top 1 matched PGF, which show in the right bars of Figure 8. We
can concluded that our method scores 517 and performs best also.
Experiments also shows some limitation of our method. For the PGFs which do not contain compound shapes, the
retrieval results are not satisfactory. For example, the query in Figure 9(a) contains a circle, a trapezoid and two
triangles, but our algorithm fails to find out efficient compound shapes. Consequently, none of the retrieved figures in
Figure 9(b) is similar to the query, though they contain similar circles, triangles and trapezoids.

(a) (b)

Figure 9 The query(a) and the top 3 retrieved results(b)

SPIE-IS&T/ Vol. 9021 90210R-10


4.4 Conclusions
In this paper, aiming to retrieve the plane geometry figures efficiently, we focus on the structure analysis to build the
structural descriptor based on the detected basic geometry shapes. The main contribution of our work is that we propose
a structure analysis method to find distinguishing compound shapes to describe the overlapped plane geometry figures. It
can reflect the structures and layouts of the figures compared with the classical low-level shape descriptors. The
matching rules address the key issues and help us complete the plane geometry figure retrieval system, which can be
applied to the mathematical teaching field. This method also provided a guide for the recognition and retrieval of multi-
contour or other complex graphics. We will add more basic shape detectors in the future to provide diversification of the
plane geometry figures and expand our dataset for commercial use.

ACKNOWLEDGMENTS
This work is supported by Natural Science Foundation of Beijing under Grant 4132033. We are deeply indebted to
the collaborators for many insightful remarks and valuable suggestions.

REFERENCES

[1] Zhang, D. and Lu, G., "Review of shape representation and description techniques," Patt. Reco. Soc., 37(1), (2004)
[2] Liao, S. X., Pawlak, M., "On image analysis by moments," IEEE Trans. Patt. Analy. and Mach., Intel., 18(3), (1996).
[3] Zhang, D. and Lu, G. "Generic fourier descriptor for shape-based image retrieval," Proc. of the IEEE ICME. 1, 425-
428(2002).
[4] Tabbone, S., Ramos, Terrades, O., Barrat, S., "Histogram of radon transform. a useval descriptor for shape retrieval,"
Proc. IEEE ICPR., 1-4(2008).
[5] Ling, H. and Jacobs, D. W., "Shape classification using the inner-distance," IEEE Trans. Patt. Analy. and
Mach.,29(2), (2007).
[6] Molhtarian, F., Abbasi, S. and Kittler, J., "Curvature scale space image in shape similarity retrieval," Multi. System,
7(6), (1999).
[7] Ballard, D. H., “Generalizing the Hough Transform to detect arbitrary shapes,” Patt. Reco. 13, 111-222(1981).
[8] Lamiroy, B., Gaucher, O. and Fritz, L., “Robust circle detection,” Proc. ICDAR, 526-530(2007).
[9] Chung, K. L., Huang, Y. H., Shen, S. M., Krylov, A. S., Yurin, D. V. Semeikina, E. V. “Efficient sampling strategy
and refinement strategy for randomized circle detection,” Patt. Reco. 45, 252-263(2012).
[10] Akinlar, C. and Topal, C., “A real-time circle detector with a false detection control,” ICASSP, 1309-1312(2012).
[11] Cuevas, E., Enciso, V. O., Wario, F., Zaldivar, D. and Cisneros, M. P., “Automatic multiple circle detection based
on artificial immune systems,” Expert Systems with Application, 39, 713-722(2012).
[12] Duda, R. O. and Hart, P. E., “Use of the Hough transformation to detect lines and curves in pictures,” Comm. ACM
15, 11-15(1972).
[13] Lin, C. and Nevatia, R., “Building detection and description from a single intensity image,” Comp. Vision Image
Under. 72(2), 101-121(1998).
[14] Nayef, N. and Breuel, T. M., "Efficient symbol retrieval by building a symbol index from a collection of line
drawings," DRR Feb 5-7 San Francisco CA U.S.A., (2013).
[15] Li, K., Lu, X., Ling, H., Liu, T., Feng, T. and Tang, Z., “Detection of overlapped quadrangles in plane geometry
figures,” ICDAR Aug 25-28 Washington DC U.S.A., (2013).
[16] Bourgeois, F. and Lassalle, J. C., “An extension of the munkres algorithm for the assignment problem to rectangular
matrices,” Comm. ACM, 14(12), 802-804(1971).
[17] Kim, W. Y. and Kim, Y. S. “A region-based shape descriptor using Zernike moments,” Sig. Proc. Image Comm. 16,
95-102(2000).

SPIE-IS&T/ Vol. 9021 90210R-11

View publication stats

You might also like