=Paper= {{Paper |id=Vol-1177/CLEF2011wn-ImageCLEF-CasanovaEt2011 |storemode=property |title=IFSC/USP at ImageCLEF 2011: Plant Identication Task |pdfUrl=https://ceur-ws.org/Vol-1177/CLEF2011wn-ImageCLEF-CasanovaEt2011.pdf |volume=Vol-1177 }} ==IFSC/USP at ImageCLEF 2011: Plant Identication Task== https://ceur-ws.org/Vol-1177/CLEF2011wn-ImageCLEF-CasanovaEt2011.pdf

IFSC/USP at ImageCLEF 2011: Plant
identification task

Dalcimar Casanova? , João Batista Florindo?? , and Odemir Martinez Bruno

USP - Universidade de São Paulo
IFSC - Instituto de Fı́sica de São Carlos, São Carlos, Brasil
bruno@ifsc.usp.br

Abstract. The leaves are one of the most important main sources used
for plant identification. Because of this the ImageCLEF 2011 proposed
a challenge based on leaf analysis for plant identification. This paper
reports the experiment results of the IFSC/USP team in participating
of this task. The main goal is investigate the performance of Complex
Network method for feature extraction and classification of plant species.
The achieved results are promising and can help the botanists in the
future.

Keywords: Complex Network, FDA, Taxonomy, Plant identification, Leaves.

1 Introduction
In the world is estimated that there are 400,000 species, of which only 270,000
have been named and identified by botanists. The Plant Taxonomy is the respon-
sible science for survey of the fauna and your consequent classification. Although
we have many researches in this field, the taxonomy of species is still a hard task.
One of reasons for this is the fact of the flowers and fruits (the main sources used
for diagnostic of characteristics) are not available for studies throughout the year,
but only at certain times. Although available throughout most of the year, the
leaves do not have sufficient visible characteristics to differentiate between many
species. Methods of computer vision can help in this point. The main idea is
extract more good characteristics of the leaves, using computer vision methods,
than traditional manual inspection.
The ImageCLEF series use this idea and propose an ongoing campaign’s
on plant identification task. The main goal of this task is provide a forum for
researchers that work on image analysis and artificial intelligence methods, share
ideas and compare their systems in order to help the taxonomic process with
leaves information.
?
Dalcimar Casanova gratefully acknowledges the financial support FAPESP (São
Paulo Research Foundation, Brazil) (2008/57313-2) for his PhD grant.
??
João Batista Florindo gratefully acknowledges the financial support CNPq (National
Council for Scientific and Technological Development) (870336/1997-5) for his PhD
grant.
Our group, which has already been working on plant identification in recent
years, accepts this challenge. In this year we use a new method of shape analysis,
based on Complex Network theory [1], to characterize the contour of leaves. This
method is based on simple measurements of Complex Networks. Although very
simple, the use of these features has shown good results in other works of shape
analysis.
The following Section 2 describes the materials and methods utilized. In
Section 3, we explain the experiments and obtained results. Finally, conclusions
are presented in Section 4.

2 Material and Methods
2.1 Database
The experiments are performed over Pl@antLeaves dataset [3]. This database
is maintained by the French project Pl@ntNet (INRIA, CIRAD, Telabotanica).
The full database contains 5436 images of 71 tree species of real-world. The
images are taken under 3 different practical conditions:
1. Scan: contains 3070 scans of leaves collected using flatbed scanners. These
images are oriented vertically along the main natural axis and with the
petiole visible.
2. Scan-like photos: contains 897 photos which look similar to the scans images.
Those images have uniform background but with some luminance variations,
optical distortions, shadows and color derivations.
3. Free natural photos: contains 1469 photos taken directly on the trees. No
acquisition protocol is used, which results in a non-uniform background,
rotated and bad-scaled images, among others problems.
Each image has an xml file associated that contain the date, type (single
leaf, single dead leaf or foliage), name of the author and GPS coordinates of the
observation among others information. But we do not use any of this information
in classification process. Just characteristics of the images are used to make our
recognition system.
The full database is split in training and test dataset. The training dataset
have 4004 images (2329 scans, 686 scan-like photos, 989 natural photos) and
de test dataset have 1432 images (741 scans, 211 scan-like photos, 480 natural
photos).

2.2 Pre-processing
All images of both test and training dataset are firstly segmented. For Scan
and Scan-like photos a simple Otsu [4] method was employed. For Free natural
photos a manual segmentation is performed where Otsu method do not have
good results.
In sequence, we apply a contour detection method to extract only contour
of leaves. We do not bother to treat open or imperfect contours, because the
method of shape analysis that we will use is robust to such problems.
2.3 Complex Network Features
To apply the Complex Network method [1] an graph G = (V, E) should be built
using the contour of the leaf. To this, each pixel of the contour S = {s1 , s2 , ..., sN }
is represented as a vertex in the network (i.e. |S| = |V |) and for each pair of
vertices an edge wij is added as their Euclidean distance:

q
wij = (xi − xj )2 + (yi − yj )2 (1)

Therefore, the network G is represented by the N × N weight matrix W
normalized into the interval [0, 1],

W
W = (2)
maxwij ∈W

A complete graph is obtained from this. So, relevant properties are extracted
from the posterior transformation of this network using a set of thresholds T =
{t1 , t2 , . . . , tL }:

aij = 0, if wij ≥ tl
ATl ∀w ∈ W (3)
aij = 1, if wij < tl

In this experiments we use T = {0.025, 0.050, 0.075, . . . , 0.925}, totaling 13
thresholds (|T | = 13). We measure, for each threshold, the mean degree and
maximum degree, given respectively by:

N
1 X
kµ = ki (4)
N i=1

kκ = max ki (5)
i

where the degree ki of a node i is the number of edges directly connected to
node, and it is defined in terms of the adjacency matrix A as:

N
X
ki = aij (6)
j=1

An normalization of the vertices degree by the number of vertices in the
network is necessary before computing these measurements. This normalization
is performed in order to reduce the influence of the network size in the computed
descriptors, and it is performed as follows:
ki
∀ki = (7)
N
Thus, a feature vector x for each leaf image is composed by 26 features (13
of kµ and 13 of kκ ). The Fig. 1 shows all process to computing these features
vectors.

Segmentation Contour
extraction

Original

Complex Network Fourier

FDA FDA

CDA + Bayes CDA + Bayes CDA + Bayes
IFSC USP run1 IFSC USP run2 IFSC USP run3

Fig. 1. Overview of runs IFSC USP

2.4 Fourier Features
The Fourier descriptors consist of the sum of the main components of the nor-
malized power spectrum. It was used here 40 frequencies. These frequencies are
then called Fourier descriptors and make up the feature vector x.
2.5 Functional Data Analysis

Functional Data Analysis (FDA) [6, 2] is a statistical approach alternative to
multivariate analysis. In FDA, a set of variables is handled as a unique entity,
more exactly, an analytical function. Such function may be obtained through
any sort of interpolation method. Thus, the function f may be calculated by:
q
X
f= αj (f )φj , (8)
j=1

where φ are the basis functions and α are the basis coefficients.
In this work, we used B-splines basis. Then, we extract features from the
Complex Network or Fourier descriptors by applying a transform to the coeffi-
cients α (alpha = x) [8, 2]. The features are represented by β(f ) and provided
by:
β(f ) = Sα(f ), (9)
where S is the Choleski decomposition of Φ matrix, whose elements are:

Φ(k, l) =< φk , φl > . (10)

The above transform turns possible the expression of the original data on
the basis functions algebraic space. In this way, it becomes a richer analysis tool
emphasizing the global relevant aspect of the original data.

3 Canonical Discriminant Analysis + Naive Bayes

With a single feature vector for each leaf we have chosen to use the Naive Bayes
as classifier. In addition, we have used a 10-fold cross validation. For g groups,
the Bayes rule assigns an object to the group i when:

P (i|x) > P (j|x), f or ∀j 6= i (11)
In this case, assuming the hypothesis of independence, we have for the ran-
dom variables:
Qn
P (i) k=1 P (xk |i)
P (i|x) = Qn (12)
k=1 P (xk )

where:
(xi −µik ) 2
1 2σ 2
P (xk |i) = p 2
e ik (13)
2πσik
being P (x|i) the probability of obtaining a particular set of features x, given
that the object belongs to the group i and P (i) is the probability a priori, that is
the probability of choosing the group i without any feature of the known object.
In addition to Naive Bayes we use the Canonical Discriminant Analysis [7].
This method aims maximize the separation between classes. Given the matrix
S, indicating the total dispersion among the feature vectors, defined as:
N
X
S= (xi − µ)(xi − µ)0 (14)
i=1

and the matrix Si indicating the dispersion of objects of Ci :
X
Si = (xi − µi )(xi − µi )0 (15)
i∈Ci

we can define the intra-class variability Sintra (indicating the combined dis-
persion within each class) and interclass variability Sinter (indicating the disper-
sion of the classes in terms of their centroids) as:
K
X
Sintra = Si (16)
i=1

K
X
Sinter = Ni (µi − µ)(µi − µ)0 (17)
i=1

where K is the number of classes, N , the number of samples, Ni , the number
of objects in class i, Ci , the set of samples of class i, µ, the global average, and
µi , the average of objects in class i. For these measures of dispersion we have
necessarily:

S = Sintra + Sinter (18)
Thus, the i-th canonical discriminant function is given by:

Zi = ai1 X1 + ai2 X2 + · · · + aip Xp (19)
where p is the number of features of the model and ai j are the elements of
the eigenvector ai = (ai1 , ai2 , . . . , aip ) of matrix C given by:
−1
C = Sinter ∗ Sintra (20)
In general a reduction in the number of features is desired. Thus, the system of
random variability of the original vector with p-original variables is approximated
by the variability of the random vector containing the k-principal components.

4 Experiments and Results
We submitted 3 different runs to the plant identification task. The main differ-
ences between those 3 runs can be seen in Fig. 1.
In the first run we apply the Complex Network method in shapes of the
leaves. In sequence the methods of Canonical Discriminant Analysis followed by
Naive Bayes classifier are employed. For that, only 10 canonical variables are
used in the Naive Bayes classifier. These 10 main components represent 99.99%
of total variance.
For the second run, the FDA method is employed over Complex Network
descriptors. The new obtained descriptors by FDA method are then used as
input to the CDA method.
The third run is exactly equal of the second run, except that, in this we use
Fourier descriptors in replace of Complex Network descriptors.
The results are showed in Table 1. We see that Complex Network descriptors
obtain best results than Fourier descriptors. You can also observe that the FDA
method make a small improving on the success rate.
Is important emphasize here that, though both methods aims improve the
quality of descriptors, each one acts in a different way on these. On Canonical
Discriminant Analysis the objective is maximize the separation of classes, while
the Functional Data Analysis aims highlight some features of the original feature
vector.

Run name No. of descriptors scan scan-like photos mean Sucess rate
IFSC USP run1 26 0.411 0.430 0.503 0.448
IFSC USP run2 26 0.562 0.402 0.523 0.496
IFSC USP run3 40 0.356 0.187 0.116 0.220
Table 1. Results for all runs in plant database.

It is important to emphasize that the method of Complex Networks do not
need a closed contour, since the method is invariant to rotation and scale, prob-
lems that we have in the dataset. We also have a good robustness against noise
and spurious contour points [1]. Perhaps due to these characteristics, the CN
method fared better than Fourier.
Is not surprising the good success rate achieved by the photo images (if
compared with scan and scan-like images). This is probably due to the manual
segmentation performed on these images.

5 Conclusion
Although we have obtained good results, they are still far from ideal. We per-
ceived as the main problem to lack of standardization of the images, especially
images of free natural photos. For these images we do not have a good generic
method to make the correct segmentation of all the images.
It is important to remember that other relevant information contained in the
associated XML was not used. Such information can help achieve better success
rates.
Is important to note that there are other diagnostic keys that can be used
to leaf identification, texture [7] and venation [5] are just some examples. These
attributes appear to contain richer information than the leaf contour. However,
in order to use of these attributes, images with higher resolution and a standard
procedure for capturing images need be used.
Thus, there are good prospects to achieve a good system of leaf identification
with the use of these variables. Such a system would be very helpful to botanists
and other professionals.
Bibliography

[1] A. R. Backes, D. Casanova, and O. M. Bruno. A complex network-based
approach for boundary shape analysis. Pattern Recognition, 42(1):54–67,
2009.
[2] J. B. Florindo, M. D. Castro, and O. M. Bruno. Enhancing multiscale fractal
descriptors using functional data analysis. International Journal of Bifurca-
tion and Chaos, 20(11):3443–3460, 2010.
[3] H. Goëau, P. Bonnet, A. Joly, N. Boujemaa, D. Barthelemy, J.-F. Molino,
P. Birnbaum, E. Mouysset, and M. Picard. The clef 2011 plant images clas-
sification task. In CLEF 2011 working notes, Amsterdam, The Netherlands,
2011.
[4] N. Otsu. A threshold selection method from grey-level histograms. IEEE
Trans. Systems, Man and Cybernetics, 9(1):62–66, 1979.
[5] R. O. Plotze, M. Falvo, J. G. Pádua, L. C. Bernacci, M. L. C. Vieira,
G. C. X. Oliveira, and O. M. Bruno. Leaf shape analysis using the mul-
tiscale minkowski fractal dimension, a new morphometric method: a study
with passiflora (passifloraceae). Canadian Journal of Botany, 83:287–301,
2005.
[6] J. O. Ramsay and B. W. Silverman. Applied Functional Data Analysis:
Methods and Case Studies. Springer-Verlag, New York, 2002.
[7] D. R. Rossatto, D. Casanova, R. M. Kolb, and O. M. Bruno. Fractal analy-
sis of leaf-texture properties as a tool for taxonomic and identification pur-
poses: a case study with species from neotropical melastomataceae (miconieae
tribe). Plant Systematics and Evolution, 291(1):103–116, 2010.
[8] F. Rossi, N. Delannay, B. Conan-Guez, and M. Verleysen. Representation of
functional data in neural networks. Neurocomputing, 64(1):183–210, 2005.