Tips & Tricks
Your SlideShare is downloading.
Like this document? Why not share!
Securing Your Active Directory Thro...
How CFOs Can Achieve Strategic Vend...
by marcus evans
Email sent successfully!
Show related SlideShares at end
Mar 03, 2014
Comment goes here.
12 hours ago
Are you sure you want to
Your message goes here
Be the first to comment
Be the first to like this
Number of Embeds
Flagged as inappropriate
Flag as inappropriate
No notes for slide
Transcript of "50120140502010"
1. International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), INTERNATIONAL JOURNAL OF COMPUTER ENGINEERING & ISSN 0976 - 6375(Online), Volume 5, Issue 2, February (2014), pp. 88-97 © IAEME TECHNOLOGY (IJCET) ISSN 0976 – 6367(Print) ISSN 0976 – 6375(Online) Volume 5, Issue 2, February (2014), pp. 88-97 © IAEME: www.iaeme.com/ijcet.asp Journal Impact Factor (2014): 4.4012 (Calculated by GISI) www.jifactor.com IJCET ©IAEME A NEW LEAF ANALYSIS AND CLUSTERING FOR TEA SPECIES IDENTIFICATION Arunpriya C. P.S.G.R Krishnammal College for Women, Coimbatore, India Antony Selvadoss Thanamani Nallamuthu Gounder Mahalingam College, Pollachi, India. ABSTRACT Leaf is an important organ of the plant. It is widely used for many purposes such as in medical field, chemical and other research purposes. Now it becomes active area for analysis of plants as most of the plant species are at the risk of extinction. Most of the leaves cannot be analyzed easily since some are not flat (e.g. succulent leaves and conifers), some does not grow above ground (e.g. bulb scales), and some does not undergo photosynthetic function (e.g. cataphylls, spines, and cotyledons).In this paper, we mainly focused on tea leaves to identify the leaf type for improving tea leaf classification. Tea leaf images are loaded from digital cameras or scanners in the system. This proposed approach consists of three phases such as preprocessing, feature extraction, selection and finally clustering of leaves. The tea leaf images are first preprocessed to remove the noise and enhanced by fuzzy denoising using Dual Tree Discrete Wavelet Transform (DT-DWT and boundary enhancement to obtain the shape of leaf accurately. In the feature extraction phase, Digital Morphological Features (DMFs) and Geometrical features are extracted and from that main features are selected. They are given to the clustering process which is done by using Fuzzy C-Means algorithm, it clearly cluster different type of tea leaves. The Fuzzy C-Means is trained by 60 tea leaves to classify them into 6 types. Experimental results proved that the proposed method clustered the tea leaves with more accuracy in less time. Thus, the proposed method achieves more accuracy in clustering the leaf type. Keywords: Leaf Recognition, Dual Tree Discrete Wavelet Transform (DT-DWT), Digital Morphological Features (DMFs), K-Means Algorithm, Fuzzy C-Means (FCM). 88
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 - 6375(Online), Volume 5, Issue 2, February (2014), pp. 88-97 © IAEME I. INTRODUCTION Plant is an important thing in our environment. It has vital use in food stuﬀ, medicine and industry. There is an difficulty arises while recognize plant species on earth. All over the world, there are currently about 310 000–420 000 known plant species, and many are still unknown yet.  At present, plant taxonomy usually adopts traditional classification method. And so far many other classification methods, such as molecule biology, morphologic anatomy, cell biology and phytochemistry, have also been used. These methods have good relation to do with biology and chemistry. However, the acquisition of needed data from plant living body or specimen directly and automatically by computer has not been implemented . Plants are basically classified according to shapes, colors and structures of their leaves and flowers [5, 6]. However, if we want to recognize the plant based on 2D images, it is difficult to analyze shapes and structures of flowers since they have complex 3D structures. On the other hand, the colors of leaves are always green; moreover, shades and the variety of changes in atmosphere and season cause the color feature having low reliability. Therefore, we decided to recognize various plants by the grey-level leaf image of plant. The leaf of plant carry useful information for classification of various plants, for example, shape, aspect ratio and texture.  In this work initially the leaf images are subjected to preprocessing, In that preprocessing the leaf image is converted into gray scale and then from it noises are removed and it is enhanced to desired level. From it several features are extracted and selection of features is made. Then they are clustered based upon their features. The paper can be organized as follows. Section II describes the related works involved regarding leaf analysis. Section III describes about the proposed methodology. Experimental results are illustrated in Section IV and Section V deals with the conclusion. II. RELATED WORKS Ji-Xiang Du et al., takes leaf database from different plants is firstly constructed. Then, a new classification method, referred to as move median centers (MMC) hypersphere classifiers, for the leaf database based on digital morphological feature is proposed. Then says that it is more robust than the one based on contour features since those significant curvature points are hard to find. Finally, the efficiency and effectiveness of the method in recognizing different plants is demonstrated by experiments. Xiao Gu et al.,  proposed a novel approach for leaf recognition by means of the result of segmentation of leaf’s skeleton based on the integration of Wavelet Transform (WT) and Gaussian interpolation. And then the classifiers, a nearest neighbor classifier (1-NN), a k -nearest neighbor classifier (k-NN) and a radial basis probabilistic neural network (RBPNN) are employed, based on Run-length Features (RF) obtained from the skeleton to identify the leaves. Ultimately, the efficiency of this approach is illustrated by several experiments. The results reveal that the skeleton can be effectively extracted from the entire leaf, and the recognition rates can be significantly improved. The evolution of the curves may be driven by image gradient information (Kichenassamy et al., 1995; Caselles et al., 1997), region information (Chan and Vese, 2001 ; Jehan-Besson et al., 2003) , or their combination (Tsai et al., 2001) . The image segments generated are enclosed by closed contours and are unsupervised which make it an ideal candidate for image post processing. These methods are more suitable for simple images and their performance degrades both in terms of segmentation and complexity, with complicated images (Paragios and Deriche, 2000). In such cases, the utilization of prior information is necessary for curve evolution methods in complicated image segmentation applications. 89
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 - 6375(Online), Volume 5, Issue 2, February (2014), pp. 88-97 © IAEME III. METHODOLOGY The tea leaf recognition method used in the proposed approach consists of three phases namely image pre processing, feature extraction and selection then finally clustering. 3.1. Image Pre-Processing The input tea leaf images undergo several processing steps as follows. Converting RGB Image to Binary Image The tea leaf image is obtained through scanners or digital cameras. An RGB image is firstly converted into a grayscale image. Equation 1 is used to convert RGB value of a pixel into its grayscale value. gray=0.2989כR+0.85870כG+0.1140כB (1) Where R, G and B corresponds to the color of the pixel, respectively. Fuzzy Denoising Using Dual Tree Discrete Wavelet Transform Here the denoising is done through Fuzzy shrinkage rule. In image denoising, where a tradeoff between noise suppression and the maintenance of actual image discontinuity should be made, solutions are required to detect important image details and accordingly adapt the degree of noise smoothing. With respect to this principle, use a fuzzy feature for single channel image denoising to enhance image information in wavelet sub-bands and then using a fuzzy membership function to shrink wavelet coefficients, consequently. Dual Tree Discrete Wavelet Transform (DT-DWT) is used as a fuzzy denoising algorithm which provides both shiftable sub-bands and good directional selectivity and low redundancy. The 2-D dual-tree discrete wavelet transform (DT-DWT) of an image is employed using two critically-sampled separable 2-D DWT’s in parallel . The advantages of the dual-tree DWT (DTDWT) over separable 2D DWT is that, it can be used to employ 2D wavelet transforms which are more selective with respect to orientation. The real part or the imaginary part of DT-DWT  produces perfect reconstruction and hence it can be employed as a stand-alone transform. Feature vector can be calculated using magnitude of sub bands. The implementation of DT-DWT is easy. An input image is decomposed by two sets of filter banks, (ܪ , ܪଵ ሻ ܽ݊݀ ܪ , ܪଵ ሻ separately and filtering the image both horizontally and vertically. Then eight sub bands are obtained: LLa, HLa , LHa , HHa , LLb , HLb , LHb and HHb. Each high-pass subband from one filter bank is combined with the corresponding subband from the other filter bank by simple linear operations: averaging or differencing. The size of every sub band is the same as that of 2D DWT at the same level.  But there are six high pass sub bands instead of three highpass sub bands at every level. The two low pass sub bands, LLb and LLa, are iteratively decomposed up to a desired level within each branch. The DT-DWT (K) can be designed in two ways to have required delays. The first is based on Farras filters and the second employs Q-shift (quarter shift) filter design. The key issue in the design of DT-DWT (K) is to obtain (approximate) shift invariance using any of the filter forms.To use a redundant transform for compression seems contradictory to the goal of compression which is to reduce whatever redundancy as much as possible. However if coefficients of a redundant transform are sparse enough, compression can even benefit from the introduced redundancy since most coefficients are nearly zero. Processing is usually result from a modification of the spatial correlation between wavelet coefficients (often caused by zeroing of small neighboring coefficients) or by using DWT. 90
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976 f 0976-6367(Print), ISSN 0976 - 6375(Online), Volume 5, Issue 2, February (2014), pp. 88-97 © IAEME After the post processing process the enhanced leaf image is obtained as a result. Boundary Enhancement The margin of a leaf is highly focused in this pre processing step. Convolving the image with a Laplacian filter of 3 × 3 spatial mask as in equation 2: 0 1 0 1 -4 1 Fig. 1 Enhanced image 0 1 0 (2) Fig. 2 Boundary Enhancement The Fig.1 shows the enhanced ima and Fig 2 shows boundary enhancement of the image proposed tea leaf recognition. To make boundary as a black curve on white background, the pixel values “0” and “1” are swapped. 3.2. Feature Extraction The proposed approach uses Digital Morphological Features (DMFs) and geometrical MFs) features. For these features, computer can obtain feature values quickly and automatically. The automatically features used for extraction in the proposed method is described as follows. A. Physiological Length The only human interfered segment of the proposed algorithm is that, the two terminals of the the main vein of the leaf should be marked through mouse click. The distance between the two terminals is the physiological length. It is represented as Lp. The red line in the Fig. 3 indicates the physiological length of a leaf. Fig. 3 Physiological length of a leaf 3: 91
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976 f 0976-6367(Print), ISSN 0976 - 6375(Online), Volume 5, Issue 2, February (2014), pp. 88-97 © IAEME B. Physiological Width Drawing a line passing through the two terminals of the main vein, infinite lines can be plotted orthogonal to that line. The count of intersection pairs between those lines and the leaf margin is also infinite. At the physiological width, the longest distance between points of those intersection pairs is defined. It is represented as Wp. As the coordinates of pixels are discrete, two lines are considered as orthogonal if their degree is 90◦ ± 0.5◦.The red line in the Fig. 4 indicates the 90 .The physiological width of a leaf. Fig. 4 Physiological width of a leaf C. Aspect Ratio The ratio of physiological length Lp to physiological width Wp is called aspect ratio and it is atio given by equation(3), (3) D. Serration Angle The teeth angle of a leaf can be defined as equation (4) using serration angle. serrati (4) Where θ is the serration angle, a is the length of first recognizable teeth from the tip of the angle and b is the breadth of first recognizable teeth from the tip of the angle. rration Fig. 5 Serration angle obtained from the tea leaf The serration angle obtained from the tea leaf using equation 4 is shown in the Fig. 5. 92
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 - 6375(Online), Volume 5, Issue 2, February (2014), pp. 88-97 © IAEME E. Segment The segment of a leaf can be defined as the ratio of first recognizable teeth in the left side from the tip of the angle ‘a’ to the first recognizable teeth in the right side from the tip of the angle ‘b’ as defined in equation (5). (5) F. Segment maximum width to Physiological length ratio The leaf is divided into 10 segments as shown in the Fig. 6. Each segment width to the physiological length ratio can be determined for all the 10 segments. G. Tip Angle The angle which is formed from the tip of the leaf to the first recognizable teeth on either side of the leaf is called tip angle. The tip angle can be calculated using the formula as defined in equation (6). (6) where θ is the tip angle, a and b are the first recognizable teeth from the tip of the angle on left and right side respectively. The tip angle formed from the tea leaf is shown in the Fig. 7. θ a Fig. 6 Segment of a leaf b Fig. 7 Tip angle obtained from the tea leaf 3.3. Feature Selection Feature selection is the process of selecting the features which are essential for its classification or clustering. While extracting the features user may extract many features. And when all the extracted features are given for clustering, it may fails in clustering. The computational complexity arises while large numbers of features are given for clustering and time consumption also becomes high. For these reasons feature selection is made and here appropriate features are selected based on its further process. 93
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 - 6375(Online), Volume 5, Issue 2, February (2014), pp. 88-97 © IAEME The clustering process here is done according to the selected features. The leaf images here are grouped based on it. In this paper several morphological and geometrical features are extracted. The geometrical features such as Physiological Length, Physiological Width, Serration Angle, Leaf Layer Segment – Layer 1, Layer 2, Layer 3, Layer 4, Layer 5, Layer 6, Layer 7, Layer 8, Layer 9 and Layer 10 maximum distance, Segment maximum width to Physiological length ratio and Tip Angle are extracted. Morphological feature of Aspect Ratio is extracted. These are the unique features by this clustering is made. Each type of leaf differs by these features. 3.4. K-means clustering To cluster the images based on the selected features k-Means Clustering is used. K-Means algorithm is one of the clustering algorithms that classify the input data points into multiple classes based on their inherent distance from each other. Suppose that the data features form a vector space and this algorithm tries to find natural clustering in them. The points are clustered around centroid µ୧ ୧ ൌ 1 … k which are obtained by minimizing the objective as given in equation (7) ଶ ܸ ൌ ∑୩ ∑୶ౠ א౩ ൫x୨ െ µ୧ ൯ ୧ୀଵ (7) where there are k clusters R ୧,୨ ൌ 1, 2, … … , k and µ୧ is the centroid or mean point of all the points x୨ אR ୧ The algorithm takes a 2 dimensional image as an input. Various steps in the algorithm are as follows: • • • • Compute the intensity distribution of the intensities. Initialize the centroids with k random intensities. Repeat the following steps until the cluster labels of the image do not change anymore. Cluster the points based on distance of their intensities from the centroid intensities which can be defined as equation (8). ଶ c ሺ୧ሻ ൌ arg min ቛx ሺ୧ሻ െ µ୨ ቛ ୨ • ሺ8ሻ Compute the new centroid for each of the clusters as in equation (9). ߤ ؔ ∑ 1 ൛ܿሺሻ ൌ ݆ൟ ݔሺሻ ୀଵ ∑ 1 ൛ܿሺሻ ൌ ݆ൟ ୀଵ ሺ9ሻ where k is a parameter of the algorithm (the number of clusters to be obtained), i iterates over the all the intensities, j iterates over all the centroids and µ୧ are the centroid intensities. . 3.5. Fuzzy C-Means Clustering The Fuzzy C-Means (FCM) clustering algorithm was first introduced by Dunn  and later was extended by Bezdek . The algorithm is an iterative clustering method that produces an optimal c partition by minimizing the weighted within group sum of squared error objective function JFCM defined in equation (10) . 94
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 - 6375(Online), Volume 5, Issue 2, February (2014), pp. 88-97 © IAEME ܬிெ ൌ ሺݑ ሻ ݀ଶ ሺݔ , ݒ ሻ ୀଵ ୀଵ ሺ10ሻ Where X = ݔଵ , ݔଶ , … , ݔ ሽ ܴ ه is the data set in the p-dimensional vector space, ݊ is the number of clusters with 2 ܿ ൏ ݊, ݑ is the degree of membership of ݔ in the ith cluster, q is a weighting exponent on each fuzzy membership, vi is the prototype of the centre of cluster i, ݀ଶ ሺݔ , ݒ ሻ is a distance measure between object ݔ and cluster centre ݒ . A solution of the object function JFCM can be obtained via an iterative process. IV. EXPERIMENTAL RESULTS The tea leaf clustering is taken in the proposed approach. For each type of tealeaf, 10 pieces of leaves from testing sets are used to test the accuracy. The dataset used in this approach is UPASI dataset. For clustering using FCM algorithm, input as features are given by those features they are clustered. Here for this UPASI dataset there are 6 types of tea leaves. Each type has nearly same feature values and they are far for the other type. TABLE I: DIFFERENT TYPES OF TEA LEAVES TESTED IN THE PROPOSED METHOD Tea Leaf Name Serration Angle Feature Values Tip Angle Feature Values TRF 1 110 to 140 35 to 50 UPASI - 3 35 to 120 35 to 75 UPASI - 9 80 to 140 28 to 46 UPASI - 10 110 to 150 45 to 55 UPASI -17 100 to 120 42 to 52 UPASI - 22 60 to 150 45 to 100 The feature values obtained for serration and tip angle are shown in table I. Likewise all other remaining 14 different features are extracted. From the above table, we can see that the range of feature values differs for each type of leaves. The performance of the proposed approach is evaluated based on the following parameters: Accuracy and Execution time. The different tea leaves taken in the proposed approach is shown in the Table II. TABLE II: DIFFERENT TYPES OF TEA LEAVES TESTED IN THE PROPOSED METHOD Tea Leaf Name Tested Samples Number of Correct Clustered TRF 1 57 54 UPASI - 3 62 52 UPASI - 9 56 51 UPASI - 10 56 46 UPASI -17 60 56 UPASI - 22 63 56 95
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 - 6375(Online), Volume 5, Issue 2, February (2014), pp. 88-97 © IAEME TABLE III: COMPARISON OF THE EXECUTION TIME Classification Time (seconds) Techniques Fuzzy C-Means 0.83 K-Means 0.74 EM 0.92 Table III reveals the execution time of the classification algorithms. The execution time of the proposed K-Means clustering approach is less when compared with EM and Mean Shift Clustering algorithm. Thus this k-means clustering is well suits for clustering of images in reduced executed time. V. CONCLUSION A new approach of tea leaf feature selection and clustering is proposed in this paper. This approach consisted of three phases namely the preprocessing, feature extraction and selection finally clustering. The image can be preprocessed using fuzzy denoising by Dual Tree Discrete Wavelet Transform (DT-DWT) and Boundary enhancement. The main features of leaf can be extracted by using morphological and geometrical feature techniques. And the related features are selected and given as an input to the clustering. FCM clustering algorithm is used as a clustering algorithm. The computer can automatically cluster 60 kinds of plants via the leaf images loaded from digital cameras or scanners. The performance of the proposed approach is evaluated based on the accuracy and execution time. The proposed algorithm produces better accuracy and takes very less time for execution. REFERENCES          Ji-Xiang, Du, Xiao-Feng Wang, D.S. Huang, Automatic plant leaves recognition system based on image processing techniques, Technical Report, Institute of Intelligent Machines, Chinese Academy of Sciences, October, 2004. Ji-Xiang Du , Xiao-Feng Wang and Guo-Jun Zhang,“ Leaf shape based plant species recognition”, Applied Mathematics and Computation 185 (2007) 883–893. Aujol, J., Aubert, G. and Blanc-Féraud, L. (2003) Wavelet-based level set evolution for classification of textured images,” IEEE Trans. Image Process., Vol. 12, No. 12, Pp. 1634–1641. National Institute for Agricultural Botany, Chrysanthemum Leaf Classification. Cambridge, 2005. Chan, T. F. and Vese, L.A. (2001) Active contours without edges, IEEE Trans. Image Process., Vol. 10, No. 2, Pp. 266–277. Im, H. Nishida, T.L Kunii, “Recognizing plant species by leaf shapes—a case study of the Acer family,” Proc. Pattern Recognition, Vol. 2, pp. 1171-1173, 1998. Z. Wang, Z. Chi, D. Feng, “Shape based leaf image retrieval,” Image Signal Process, Vol. 150, no. 1, pp.34-43, 2003 Im, H. Nishida, T.L Kunii, “Recognizing plant species by leaf shapes—a case study of the Acer family,” Proc. Pattern Recognition, Vol. 2, pp. 1171-1173, 1998. Z. Wang, Z. Chi, D. Feng, “Fuzzy integral for leaf image retrieval,” Proc. Fuzzy Systems, Vol. 1, pp.372-377, 2002. T. Saitoh, T. Kaneko, “Automatic recognition of wild flowers,” Proc. Pattern Recognition, Vol. 2, pp. 507 -510, 2000. 96
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 - 6375(Online), Volume 5, Issue 2, February (2014), pp. 88-97 © IAEME  Warren, “Automated leaf shape description for variety testing in chrysanthemums,” Proc. Image Processing and Its Applications, Vol. 2, pp. 497-501, 1997.  Z. Wang, Z. Chi, D. Feng, “Shape based leaf image retrieval,” Image Signal Process, Vol.150, no. 1, pp.34-43, 2003.  N.Valliammal and S.N.Geethalakshmi, “Automatic Recognition System Using Preferential Image Segmentation For Leaf And Flower Images”, Computer Science & Engineering: An International Journal (CSEIJ), Vol.1, No.4, October 2011.  Xiao Gu, Ji-Xiang Du and Xiao-Feng Wang, “Leaf Recognition Based on the Combination of Wavelet Transform and Gaussian Interpolation”, Advances in Intelligent Computing, Vol. 3644/2005, pp. 253-262, 2005..  Chen, Y. (2002) Using prior shapes in geometric active contours in a variational framework, Int. J. Comput. Vis., Vol. 50, No. 3, Pp. 315–328.  Suman Tatiraju and Avi Mehta, “Image Segmentation using k-means clustering, EM and Normalized Cuts.”.  M. J. L. Orr, “Introduction to radial basis function networks”, Technical report, Center for Cognitive Science, University of Edinburgh, 1996  Arokia Priya, S.PNarote, and A.V.Patil, “Dual Tree Wavelet Transforms in Image Compression”.  M. Sepasian, W. Balachandran and C. Mares, “Image Enhancement for Fingerprint Minutiae-Based Algorithms Using CLAHE, Standard Deviation Analysis and Sliding Neighborhood”, Proceedings of the World Congress on Engineering and Computer Science 2008 WCECS 2008, October 22 - 24, 2008, San Francisco, USA.  R.Sharmila, R. Uma, “A New Approach To Image Contrast Enhancement using Weighted Threshold Histogram Equalization with Improved Switching Median Filter”, International Journal Of Advanced Engineering Sciences and Technologies Vol No. 7, No. 2, pp. 206 – 211, 2011.  Dunn, J.C.: A Fuzzy Relative of the ISODATA Process and its Use in Detecting Compact Well Separated Clusters. Journal of Cybernetics, Vol. 3, 1974, pp. 32–57.  Bezdek, J.C.: Pattern Recognition with Fuzzy Objective Function Algorithms. New York: Plenum Press, 1981.  Abhishek Choubey, Omprakash Firke and Bahgwan Swaroop Sharma, “Rotation and Illumination Invariant Image Retrieval using Texture Features”, International Journal of Electronics and Communication Engineering & Technology (IJECET), Volume 3, Issue 2, 2013, pp. 48 - 55, ISSN Print: 0976- 6464, ISSN Online: 0976 –6472.  C. S. Sumathi and A.V. Senthil Kumar, “Identification of Arable and Tree Crops by Edge and Texture Fusion Techniques”, International Journal of Computer Engineering & Technology (IJCET), Volume 4, Issue 6, 2013, pp. 167 - 174, ISSN Print: 0976 – 6367, ISSN Online: 0976 – 6375.  Ankit Vidyarthi and Ankita Kansal, “A Survey Report on Digital Images Segmentation Algorithms”, International Journal of Computer Engineering & Technology (IJCET), Volume 3, Issue 2, 2012, pp. 85 - 91, ISSN Print: 0976 – 6367, ISSN Online: 0976 – 6375.  Garima Agarwal, Rekha Nair and Pravin Shrinath, “A Review of Plant Leaf Classification Features and Techniques”, International Journal of Computer Engineering & Technology (IJCET), Volume 4, Issue 5, 2013, pp. 204 - 216, ISSN Print: 0976 – 6367, ISSN Online: 0976 – 6375. 97