(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 4, July 2010
Plant Classification Based on Leaf Recognition
Abdolvahab Ehsanirad
Department of Computer ScienceIslamic Azad University, Minoodasht Branch, IranEmail: vahab61@gmail.com
Abstract
—
Current study used the image processing techniquesin order to classification of plants based on leaves recognition.Two methods called the Gray-Level Co-occurrence matrix(GLCM) and Principal Component Analysis (PCA) algorithmshave been applied to extract the leaves texture features. Toclassify 13 kinds of plants with 65 new or deformed leaves as testimages, the Algorithms are trained by 390 leaves. The findingsindicate that the accuracy of PCA method with 98% come out tobe more efficiency compare to the GLCM method with78%accuracy.
Keywords - Classification, GLCM, PCA, Feature Extraction.
I.I
NTRODUCTION
Leaf recognition is a pattern recognition task performed specifically on leaves. It can be described asclassifying a leaf either "known" or "unknown", after comparing it with stored known leaves. It is also desirable tohave a system that has the ability of learning to recognizeunknown leaves.Computational models of leaf recognition must addressseveral difficult problems. This difficulty arises from the factthat leaves must be represented in a way that best utilizes theavailable leaf information to distinguish a particular leaf fromall other leaves.Compared with other methods, such as cell and molecule biology methods, classification based on leaf image is the firstchoice for plant classification. Samplingleaves and photogenythem are low-cost and convenient. One can easily transfer theleaf image to a computer and a computer can extract featuresautomatically in image processing techniques. Some systemsemploy descriptions used by botanists. But it is noteasy toextract and transfer those features to a computer automatically.It is difficult job to tell the just one algorithm alone is the bestand successful at recognizing any and all variation of the sameobject. And it is more difficult to tell the samealgorithm to beable to differentiate between different objects. Many researchhas done for the leaf classification with some texture featureextraction methods [3,9,10,7].II.L
EAF CLASSIFICATION PROCESS METHOD
The conventional method of leaf classification involvestwo main steps. The first step is obtaining a priori knowledgeof each class to be recognized. Normally this knowledgeencompasses some sets of texture feature of one or all of theclasses. Once the knowledge is available and texture feature of the observed image are extracted, then classificationtechniques, for example nearest neighbors and decision trees,can be used to make the decision [5],thatis the second step.Such a procedure is illustrated in Figure 1, the tasks thattexture classification has been applied to include theclassification of plant leaves images [2].Currently there are a huge number of texture feature extractionmethods available and most of the methods are associated withtunable parameters. It is difficult to find the most suitablefeature extraction methods and their optimal parameters for a particular task. In addition, performance of classificationmethods also depends upon the problems, which makesselecting an optimal "feature extraction + classification"combination a difficult assignment.
Figure 1. Conventional Plant Classification Process
III.F
EATURE EXTRACTION
Different features are chosen to describe different properties of the leaves. Some leaves are with very distinctiveshape, some have very distinctive texture patterns, and someare characterized by a combination of these properties.IV.T
EXTURE ANALYSIS
Texture analysis mainly aims to computationallyrepresent an intuitive perception of texture and to facilitateautomatic processing of the texture information for artificialvision systems. The process of texture analysis usually produces kind of numeric descriptions of the texture, calledtexture features. The process of computing the texture featureis known as feature extraction.
OutputClassesInputimage
Texture FeatureExtraction
Classification
78http://sites.google.com/site/ijcsis/ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 4, July 2010
There are an enormous number of texture analysis methodsunder this category although none predominates. Methods thatwe used for classification will describe here.V.G
RAY
-L
EVEL
C
O
-
OCUURRENCE
M
ATRIX
This method was first proposed by Haralick in 1973 andstill is one of the most popular means of texture analysis [8].The key concept of this method is generating features based ongray level co-occurrence matrices (GLCM). The matrices aredesigned to measure the spatial relationships betweenpixels.The method is based on the belief that texture information iscontained in such relationships.Co-occurrence features are obtained from a gray-level co-occurrence matrix. We used 22 features that extracted fromGLCM matrix in our paper [8,4,1].VI.T
EXTURAL FEATURES EXTRACTED FROM GRAY
-
LEVELCO
-
OCUURENCE MATRICES
Our initial assumption in characterizing image texture isthat all the texture information is contained in the gray-levelCo-occurrence matrices. Hence all the textural features hereare extracted from these gray-level Co-occurrence matrices.The equations which define a set of 22measures of texturalfeatures are given in this paper. Some GLCM Extractedtextural features are illustrated in Table 1 for two different leaf images.
Table 1. GLCM Extracted textural features for two different leaf images.
VII.L
EAF CLASSIFICATION USING
PCAIn this study, wehave followed the method which was proposed by M. Turk and A. Pentland [6] inorder to develop aleaves classification system based on the eigenspace approach.They argued that, if a multitude of leaf images can bereconstructed by weighted sum of a small collection of characteristic features or eigenpictures, perhaps an efficient wayto learn and recognize leaves would be to build up thecharacteristic features by experience over time and recognize particular leaf by comparing the feature weights needed toapproximately reconstruct them with the weights associated withknown leaves. Therefore, each leaf is characterized by a small setof feature or eigenpicture weights needed to describe andreconstruct them. This is an extremely compact representationwhen compared with the images themselves.VIII.E
XPERIMENTAL RESULTS AND DISCUSSINONS
The experiment is designed to illustrate the performanceof two feature extraction methods, GLCM and PCAalgorithms for plant leaves classification purpose.The GLCM is a tabulation of how often differentcombinations of pixel brightness values (grey levels) occur inan image. The classification steps are illustrated in Figure 2.In the first experiment after changing the color image to gray-level image with using of the GLCM texture feature extractionweextracted the 22features [8,4,1] of each leaf images.
Figure 2. Classification Steps in GLCM method.Table 2. Somefeatures extracted from some chosen leaf image of eachleaves classes in (d=1) and degree 0°.
(a)(b)
some texture Features extracted from Leaf image (a)
AngleAutocorrelationEntropyContrastCorrelationHomogeneity0°45.57481.43110.31840.96380.614445°45.27991.49280.44580.94960.606090°45.61901.38860.23010.97380.6166135°45.29321.47160.41920.95260.6074
some texture Features extracted from Leaf image (b)
AngleAutocorrelationEntropyContrastCorrelationHomogeneity0°54.85400.89720.43610.94010.821445°54.73710.91320.44380.93960.819990°54.97970.84050.18450.97470.8267135°54.66100.93100.59610.91890.8172
Gray LevelCo-Occurrence matrixFeatureExtractionClassification
FeaturesSample leaf from leaves Classes
Class 1Class 2Class 3Class 4Class 5Class 6
Autocorrelation
55.2737349.5699460.2922554.3123445.2522250.65949
Contrast
0.3322780.3680380.3224680.3066460.3101270.333228
Correlation
0.9552640.9614780.870630.9562960.9618980.959497
Dissimilarity
0.1268990.1743670.1231010.1386080.1724680.151266
Energy
0.6991920.5280280.8282670.6469480.3874190.549963
Entropy
0.8171561.2861970.5743641.0344761.527481.168368
79http://sites.google.com/site/ijcsis/ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 4, July 2010
Wehave tried the GLCM method with Distance 1 (d=1) anddegree 0°, Distance 1 (d=1) anddegree 45°, Distance 1 (d=1)and degree 90° and Distance 1 (d=1) and degree 135°. The performance accuracy of each one is shown in Table 3.In ourexperience, GLCM method in leaf recognition for thedegrees0° and 90° gave the same accuracy and same result.Here the poor result is in the 45° degree. Because any changesin the neighboring distance or the neighboring degree it willchange the value of extracted texture feature.
78
.
46
GLCM
49
.
23
GLCM
78
.
46
GLCM
70
.
76
GLCM
98
.
46
PCA
020406080100
Figure 3. PCAand GLCMaccuracy chart in different degrees
The GLCM method is very sensitive for the any changes in theimages such rotation, scale and
etc
. In (Tables 2)you can seethe somein extracted features in neighborhood degree0.Thecomputationtime for GLCM method is less andrecognition of this method is very fast.PCA method mostly using for the face recognition purpose butwetried as leaf recognition. In PCA also image should bechange to gray level that can reduce the image dimension. Inourexperience the PCA method gave the efficient performance and very good result. It was the just one wrongrecognition out of 65 test image in ourtest. But the test speedis not much goodand computation time is high for recognizingone testimage. Compare with GLCM it's very slow but the performance of PCA method is efficient (Figure 3).IX.D
ATABASE
The database used in ourexperiment is collected by our self. Wepluck the leaf from the plant in the fields near our campusand around University of Mysore,which consists of intact and fresh leaf images in different rotation for 13 plantspecies class and constructed by our self. We taken 390images as training set andeach plant class contains the 30 leaf imagesin different degree of rotation and different leaf images. The test set contains the 65 of deformed and new leaf images and for each class has 5 leaf imagesfor test.Thesample dataset of leaf images and related classes are illustratedin Figure 4.
Figure 4. The sample dataset of leaf images and related classes
X.C
ONCLUSION
In this study, the classification based on the recognizingthe leaves images with extracted texture features was proposedand performed. The texture features have been extractedwithusing theGLCMand the PCA algorithms,on the 390image indataset and with 65 deformed or new leaf images for test. Inaddition, different degrees for the GLCM method were usedand it was found out to be more efficient in the degree 0° by78.46 % accuracy. Therefore, it wasspecified that the GLCMis very sensitive in any changes forimages such as deformingor giving the new leaf image as a test. In addition, the PCAmethod comes out to be more efficient compare tothe GLCMmethod by 98.46 % accuracy. Considering the time of
Average recognition rate (%)degree78.460°49.2345°78.4690°70.76135°98.46PCATable 3.The performance of GLCM method in different degrees withneighborhood distance 1 and performance of PCA method.
80http://sites.google.com/site/ijcsis/ISSN 1947-5500
Search History:
Result 00 of 00
00 results for result for
p.
Plant Classification Based on Leaf Recognition
Current study used the image processing techniques in order to classification of plants based on leaves recognition. Two methods called the Gray-Level Co-occurrence matrix (GLCM) and Principal Component Analysis (PCA) algorithms have been applied to extract…
(More)
2.0K
Reads
11
Readcasts
2
Embed Views
More From This User
Notes
Load more
