Classification of Diseased Plant Leaves using Feed Forward Neural Network and Learning Vector Quantization Algorithm

Classification of Diseased Plant Leaves using Feed Forward Neural Network and Learning Vector Quantization Algorithm

K. Muthukannan

Associate Professor, Dept of ECE, Einstein College of

Engineering,Tirunelveli,India

P. Latha

Associate Professor, Dept of CSE, Government College of

Engineering,Tirunelveli,India

1. Pon Selvi PG scholar, Dept of ECE,

   Einstein College of Engineering,Tirunelveli,India

   1. Nisha PG scholar, Dept of ECE,

      Einstein College of Engineering,Tirunelveli,India

      Abstract- Agriculture plays important role in human civilization. Research in agriculture domain is aimed towards increase the productivity and food quality at reduced expenditure with increased profit. But due to plant diseases, the quality of the agricultural products may be degraded. Disease in plants, that interrupts its vital functions such as photosynthesis, fertilization etc. To avoid this, the detected spot diseases in leaves are classified based on the diseased leaf types using feed forward neural network and learning vector quantization algorithm by processing the set of shape and texture features from the affected leaf portion. By this approach one can detect the diseased leaf variety and thus can take necessary steps in time to minimize the loss of production. The simulation results of classification shows the effectiveness of the proposed system. With the help of this work, a machine learning based system can be formed for the improvement of the crop quality in the Indian Economy.

      Key words – classification, feed forward neural network, learning vector quantization, performance evaluation, accuracy, precision, recall ratio, F_measure.

      1. INTRODUCTION

         Agriculture is the mother of all cultures. It plays a vital role in the development of human civilization. But plant leaf diseases can harm the crops and there may be economic loss in crops. Without knowing about the diseases affected in the plant, the farmers are using excessive pesticides for the plant disease treatment. Research in agriculture domain is aimed towards increase the quality and quantity of the product at less expenditure with more profit. The quality of the agricultural product may be degraded due to plant diseases. These diseases are caused by pathogens viz.., fungi, bacteria and viruses. Therefore, to detect and classify the plant disease in early stage is a significant task. Farmers require constant monitoring of experts which might be prohibitively expensive and time consuming.

         Depending on the applications, many systems have been proposed to solve or at least to reduce the problems, by making use of image processing and some automatic classification tools.

         Al-Bashish et.al, developed K-means-based segmentation and neural networks based classification for plant leaf disease classification [1].The proposed masking technique is a robust technique for the detection of plant leaf diseases. The developed algorithms efficiency can successfully detect and classify the examined diseases [2]. Arivazhagan S, Newlin Shebia R, developed automatically detect the symptoms of diseases as soon as they appear on plant leaves[3]. The automated pixel wise classification used to classify the sugar beet leaf diseases such as k- nearest neighbour and bayes classification technique [4]. The ANN classification is calculated by giving different type of features i.e. size, color, proximity and average centroid distance [5]. The texture features are extracted using Run length Matrix. These extracted features are then used for classification purpose using ANN classifier [6]. Plant leaf images are classified based on two different shape modelling techniques, the first based on the Moments-Invariant (M-I) model and the second on the Centroid-Radii (C-R) model [7]. Phadikar S proposed an automated system has been developed to classify the leaf brown spot and the leaf blast diseases of rice plant based on the morphological changes of the plants caused by the diseases[10]. Sanjeev S Sannakki., et al, in paper titled Diagnosis and Classification of Grape Leaf Diseases Using Neural Networks, proposed Feed forward back propagation neural network was trained for classification [15]. The color feature extraction from RGB color model where the RGB pixel color indices have been extracted from the identified region of interest [16].

      2. PROPOSED METHOD

         The proposed system includes four modules to classify the diseased plant leaves. The modules are Image collection, Feature extraction, diseased plant leaves classification and Performance evaluation.

         The below Figure 1 explains the proposed system for the diseased leaf classification. The step by step process is explained below.

         1. The disease affected leaf images are collected from various agricultural fields.

         2. Extract the texture and shape feature from diseased leaf.

         3. Classification of diseased plant leaves processed by the following algorithms.

            • Feed forward neural network algorithm

            • Learning vector quantization algorithm

         4. Performance metrics for diseased plant leaves classification calculated. There are four performance metrics are calculated to evaluate the performance of the classifier such as Accuracy, Precision, Recall ratio, F_measure.

         Fig.2. Input Image Samples: a) Bean leaf b) Bitter gourd leaf

         B. Feature Extraction

         Feature extraction is a superior form of dimensionality reduction. When the input data to an algorithm is too large to be processed and it is suspected to be notoriously redundant then the input data will be transformed into a reduced representation set of features (also named features vector). Transforming the input data into the set of features is called feature extraction.

         The data for this work have been collected from

         118 plant leaves (Bean leaf- 63, Bitter gourd- 55) which have symptoms of leaf disease. The data have been standardized so as to be error free in nature. Here various texture and shape features of an image extracted by the following formulas.

         Input image (Leaf samples – Beans & Bitter gourd)

         Extracted features from leaf samples

         1

         = , 2

         ,=0

         1

         = ,( )2

         .=0

         1 ( )( )

         Classification algorithms

         Classification algorithms

         = [

         ,=0

         1

         ,

         ,]

         1. Feed forward neural network algorithm

         2. Learning Vector Quantization

         = 1 + | |

         ,=0

         The extracted feature values of some sample diseased leaf images are given below

         TABLE I. EXTRACTED FEATURES FOR BITTER GOURD LEAF SAMPLES

         Performance Evaluation

         Classify Diseased Plant Leaves

         Performance Evaluation

         Classify Diseased Plant Leaves

         Samples Features	Image 1	Image 2	Image 3	Image 4
         Contrast	0.05810	0.04209	0.07349	0.04033
         Homogeneity	0.97115	0.97905	0.96732	0.98130
         Energy	0.21825	0.38706	0.40597	0.52695
         Correlation	0.99066	0.98602	0.98225	0.99314
         Area	39966		53973	65031	46919

         Samples Features	Image 1	Image 2	Image 3	Image 4
         Contrast	0.05810	0.04209	0.07349	0.04033
         Homogeneity	0.97115	0.97905	0.96732	0.98130
         Energy	0.21825	0.38706	0.40597	0.52695
         Correlation	0.99066	0.98602	0.98225	0.99314
         Area	39966	53973	65031	46919

         Fig 1. Workflow model for the proposed system

         A. Image collection

         The various plant leaf images are collected directly from the agricultural field using digital camera (20 Megapixel). The white background is set to take the flash of each leaf images to provide better result. In this two different agricultural plant leaves are considered. (i.e.) Bean and Bitter gourd leaf. The input sample images are shown below

         Samples Features	Image 1	Image 2	Image 3	Image 4
         Contrast	0.06643	0.11205	0.09106	0.02866
         Homogeneity	0.96739	0.94487	0.95507	0.98566
         Energy	0.20725	0.14031	0.15453	0.37750
         Correlation	0.98636	0.98330	0.98627	0.98826
         Area	48289	16000	40890	47016

         Samples Features	Image 1	Image 2	Image 3	Image 4
         Contrast	0.06643	0.11205	0.09106	0.02866
         Homogeneity	0.96739	0.94487	0.95507	0.98566
         Energy	0.20725	0.14031	0.15453	0.37750
         Correlation	0.98636	0.98330	0.98627	0.98826
         Area	48289	16000	40890	47016

         TABLE II. EXTRACTED FEATURES FOR BEAN LEAF SAMPLES

         Thus the shape and texture features of 118 diseased

         of features in which the essential information content of the input data is concentrated.

         2. The second step is the classification where the feature domains are assigned to individual classes.

         D. Performance analysis

         To measure the quality of the classified diseased leaf images the performance is analysed by using four parameters, which includes Accuracy (AC), Recall ratio, Precision and F_Measure.

         1. Accuracy

            The accuracy (AC) is the proportion of the total number of predictions that were correct. It is determined using the equation

            leaf samples are extracted and given as the input to the

            Accuracy (AC) = +

            +++

            (1)

            classifier.

            C. Image classification

            In this proposed method the classification techniques are used to classify the diseased plant

         2. Recall ratio

            The recall or true positive rate (TP) is the proportion of positive cases that were correctly identified, as calculated using the equation

            leaves. Here artificial neural networking (ANN)

            Recall ratio =

            +

            (2)

            technique is used. The ANN classification techniques as Feed forward neural network algorithm (FFNN), Learning vector quantization (LVQ) techniques are used.

         3. Precision

         Precision (P) is the proportion of the predicted positive cases that were correct, as calculated using the equation

         1) Feed forward neural network algorithm

         Precision (P) =

         +

         4) F_Measure

         (3)

         Artificial neural networks are the very versatile tools and have been widely used to tackle many issues.

         The F-Measure computes some average of the information retrieval precision and recall metrics.

         Feed-forward neural networks (FFNN) is one of the popular structures among artificial neural networks. These efficient networks are widely used to solve complex problems by modelling complex input-output relationships.

         = (2+1)

         2

         2

         Where,

         tp is number of correct classified bean leaves tn is number of misclassified bean leaves

         (4)

         1. Learning vector quantization algorithm

            Learning Vector Quantization (LVQ) is a supervised version of vector quantization that can be used when we have labeled input data.

            Fig 3. Block diagram for LVQ

            It is a two stage process a SOM followed by LVQ: This is particularly useful for pattern classification problems.

            1. The first step is feature selection the unsupervised identification of a reasonably small set

         fp is number of correct classified bitter gourd leaves

         fn is number of misclassified bitter gourd leaves

      3. EXPERIMENTAL RESULTS AND DISCUSSION

         The classification of diseased plant leaves performed by using feed forward neural network and learning vector quantization techniques which have been analyzed for the 118 input leaf images. From the confusion matrix of the respective classifier the performance metrics evaluated.

         Here four performance metrics had been calculated to measure the efficiency of the classification results.

         (i.e.) Accuracy, Precision, Recall ratio, F_measure. Figure 3 shows the confusion matrix for feed forward neural network classification for 118 input leaf samples. Here 63 bean and 55 bitter gourd

         samples are considered to extract the feature.

         Fig 4. Confusion matrix for FFNN

         From the above confusion matrix, 58 bean leaf samples are correctly classified and 5 bean leaf samples are misclassified. The correct classification rate for beans samples 92.1%. For bitter gourd, 48 samples are correctly classified out of 55.The correct classification rate for bitter gourd leaves is 89.1%.

         The overall system accuracy for FFNN classification for the above leaf samples are 90.7% and the error rate of the system is 9.3%

         .

         Fig. 5. Confusion matrix for LVQ

         From Figure 4, learning vector quantization classification result was explained. From LVQ classification, 50 bean leaf samples are correctly classified out of 63 samples. The correct classification rate for beans samples 79.4%. For bitter gourd only 18 samples are correctly classified out of 55.The correct classification for bitter gourd leaves is 32.7%.

         The overall system accuracy for the above learning vector quantization is 57.6% and the error rate of the system is 42.4%.

         The performance metrics for the above classification techniques using Accuracy, Recall ratio, Precision and F_Measure tabulation is given below

         TABLE III. PERFORMANCE EVALUATION FOR CLASSIFICATION

         Classifier Performance metrics	FFNN	LVQ
         Accuracy	0.8983	0.5762
         Precision	0.8923	0.5747
         Recallratio	0.9206	0.7936
         F_Measure	0.9062	0.6666

         1

         0.9/p>

         0.8

         0.7

         0.6

         0.5

         0.4

         0.3

         0.2

         0.1

         0

         FFNN

         LVQ

         1

         0.9

         0.8

         0.7

         0.6

         0.5

         0.4

         0.3

         0.2

         0.1

         0

         FFNN

         LVQ

         Fig 6. Performance analysis chart for classification

         From figure6, the performance analysis chart reveals that the accuracy of FFNN is higher than

         Learning Vector Quantization (LVQ). This indicates the feed forward neural network classification approach is better based on these parameters.

      4. CONCLUSION AND FUTUREWORK

In this paper, the neural network techniques such as feed forward neural network (FFNN) and learning vector quantization (LVQ) were tested for two different diseased leaf image classifications such as bean and bitter gourd leaves. The performance is measured using classification parameters such as Accuracy, Precision, Recall ratio and F_Measure. With these four parameters the performance is analyzed and based on the analysis the FFNN classification approach provides better result.

The feed forward neural network classification provides better result than learning vector quantization. The feed forward neural network provides 89.8% classification accuracy for bean and bitter gourd diseased leaves classification where as LVQ network provides 57.6% better result.

The future scope of this work will focus on developing hybrid algorithms for achieve better classification result. Genetic algorithm is combined with these neural network classification algorithms for the purpose of weight bias optimization to improve the accuracy of the classification result.

ACKNOWLEDGMENT

We sincerely thank our Principal Dr.K.Ramar and management of Einstein College of Engineering for providing full support and encouragement for preparing this paper.

REFERENCES

[1]. Al-Bashish, D., Braik M, and Bani-Ahmad S (2011), Detection and classification of leaf diseases using K-means-based segmentation and neural networks based classification Information Technology Journal, Vol 10, pp 267-275.

[2]. Al-Hiary, H., S. Bani-Ahmad, M. Reyalat, M. Braik, and Z.AlRahamneh. (2011). Fast and accurate detection and classification of plant diseases. International Journal of Computer Applications, Vol 17 , No 1,pp 31-38.

[3]. Arivazhagan S, Newlin Shebia R (2013), Detection of unhealthy region of plant leaves and classification of plant leaf diseases using texture features, Agricultural Engineering Institute : CIGR journal, volume 15, No.1.

[4]. Bauer, S. D., F. Korc, W. Forstner (2011). The potential of automatic methods of classification to identify leaf diseases from multi spectral images. Precision Agriculture, Vol 12,pp 361-377.

[5]. Hrishikesh, Kanjalkar P and Prof. Lokhande S (2013), Detection and Classification of Plant Leaf Diseases using ANN, International Journal of Scientific & Engineering Research, ISSN: 2229 5518.

[6]. Jagadeesh Devdas Pujari, Rajesh Yakkundimath and Abdulmunaf Syedhusain Byadgi (2013), Grading and Classification of Anthracnose Fungal Disease of Fruits based

on Statistical Texture Features, International Journal of Advanced Science and Technology.

[7]. Jyotismita Chaki and Ranjan Parekh (2011), Plant Leaf Recognition using Shape based Features and Neural Network Classifiers, International Journal of Advanced Computer Science and Applications.

[8]. Lu D and Weng Q (2007), A Survey of Image Classification Methods and Techniques for Improving Classification Performance, International Journal of Remote Sensing,pp 823870.

[9]. Muthukannan K et al.,(2015),Survey on Classification Techniques for Plant Leaf Disease Classification,International Journal of Research and Scientific Innovation, Vol 2, pp:15-18

[10]. Phadikar S (2012), Classification of Rice Leaf Diseases Based on Morphological Changes, International journal of Information and Electronics Engineering.

[11]. Pooja kamavisdar, sonam saluja and sonu agrawal (2013), A survey on image classification Approaches and techniques, international journal of advanced research in computer and communication engineering, ISSN: 2319-5940.

[12]. Rafael C.Gonzalez and Richard E.Woods, Digital Image Processing, Pearson Education, Third Edition.

[13]. Revathi P, Hemalatha M (2012), Classification of Cotton Leaf Spot Diseases Using Image Processing Edge Detection Techniques, International Conference on Emerging Trends in Science, Engineering and Technology.

[14]. Robert M Haralick (1973), Textural Features for Image Classification, IEEE transactions on Systems, Vol.SMC-3, No.6, pp 610-621.

[15]. Sanjeev S Sannakki (2013), Diagnosis and Classification Of Grape Leaf Diseases Using Neural Networks, 4th ICCCNT.

[16]. Suhaili Beeran kutty, Noor ezan Abdullah (2013), Classification of Watermelon Leaf Diseases Using Neural Network Analysis, IEEE Business Engineering and Industrial Applications Colloquium (BEIAC).