Machine Vision Based Paddy Leaf Recognition System

Size: px

Start display at page:

Download "Machine Vision Based Paddy Leaf Recognition System"

Marian Cobb
5 years ago
Views:

Volume 119 No. 15 2018, 3391-3396 ISSN: 1314-3395 (on-line version) url: http://www.acadpubl.eu/hub/ http://www.acadpubl.eu/hub/ Machine Vision Based Paddy Leaf Recognition System 1T.

Srinivasan 1,3Department of ECE, Srinivasa Ramanujan Centre, SASTRA Deemed University, Kumbakonam, 2Department of E&I, School of EEE, SASTRA Deemed University, Thanjavur.

1 Volume 119 No , ISSN: (on-line version) url: Machine Vision Based Paddy Leaf Recognition System 1T. Gayathri Devi, 2P.Neelamegam, and 3A. Srinivasan 1,3Department of ECE, Srinivasa Ramanujan Centre, SASTRA Deemed University, Kumbakonam, 2Department of E&I, School of EEE, SASTRA Deemed University, Thanjavur. 1 id: Abstract Paddy is the necessary food in the agriculture field. But the growth of paddy is limited by disease like Blast; Bacterial Leaf Blight; Rice tungro etc. If suitable disease diagnosis is taken at primary stage it will make a loss in paddy production. A novel method that is machine vision system for disease detection of paddy leaf is used to reduce large scale defects is described in this work. Using this method the early detection of diseases can be done and is possible to minimize the productivity loss by applying well suitable methods. Initially a paddy leaf image is captured and then enhancement processes are performed. The resultant images are given as input to classifier for classifying diseases such as Bacterial Blight, Leaf Blast and Brown Spot. Keyword: Artificial Neural Network, Field images, GLCM, FCMC, HSV Pattern recognition. 1. Introduction Indian economy mainly depends on the agricultural productivity. To assess the development of any country agriculture is the major thing. Diseases are the principal cause of yield loss of paddy production. Farmers which are lack of training cannot be able to implement the technological aids like robots and sensors. But general technological applications like machine vision, image processing and virtual environment are easy to implement in the agriculture. A widely used method to automate the detection of disease is to use computer vision. Machine vision provides the optimal solution to extract and recognize the plants based on different features automatically. In this work, three types of diseases namely Bacterial Blight, Leaf Blast and Brown spot are taken and figure.1 shows some sample input images. Figure.1 Sample of input images 2. Literature Survey The author [1] has described a stereo leaf image recognition system using gray level co- occurrence matrix. By using this method accuracy of 83.3% is obtained A leaf recognition using Zernike moments is proposed in [2]. In this method the leaves are identified using color and texture features. To grade the images using neural networks is proposed in [3]. Accuracy, error rate and sensitivity reviews the performance. In [4], the plant leaf classification. using texture feature. A color image clustering [5] that focuses on color as feature and considers is proposed. A system is developed to recognize and detect plants using various features, In[6] is it is discussed on a dial basis probabilistic neural network (RBPNN). In [7] have taken into an account of various merits of image processing for recognition of plant diseases using features like color and texture. It is proposed that species detection using images of flowers[8]. In [9] probabilistic neural network is used to analyze features. Apple classifier using neural 3391

network in [10] [11] [12] utilized to grade quality of apples, oil palm fruits. Image retrieval method is described in [13][14].

Leaf veins are extracted in [17] to define features for plant species recognition. The results show that combined approach is good.

And also explains the results and discussion and finally conclusion. 3.

2 network in [10] [11] [12] utilized to grade quality of apples, oil palm fruits. Image retrieval method is described in [13][14]. LiyingZheng, et al, developed a system to extract green vegetation in color images in [15]. A system is designed and to recognize citrus plant diseases [16] using colour co-occurrence method. Leaf veins are extracted in [17] to define features for plant species recognition. The results show that combined approach is good. This paper is organized to describe the literature survey, proposed methodology which includes image acquisition, preprocessing, feature extraction and classification. And also explains the results and discussion and finally conclusion. 3. Proposed Methodology The proposed method has the following steps Images of paddy leaves, affected by different diseases are collected and categorized. Statistical features are extracted from image samples using GLCM (Grey level cooccurrence matrix). The Artificial Neural Network is used to classify the images. The Flow diagram of proposed method is depicted in figure.2 for contrast enhancement the image preprocessing and histogram equalization is performed. Image Segmentation Fuzzy C-Means Clustering (FCMC) Algorithm is used for segmenting the preprocessed image. This clustering method allows all data points which is present more than one group. Each cluster expresses the level of individual data points present in the cluster. In this clustering is performed in iterative manner. Fuzzy C-Means Clustering (FCMC) Algorithm has the following steps Step1: Specifying the number of clusters. Step2: Separates the data into C Fuzzy clusters. Step3: Computing the mean squared error Step4: Updating the membership function. Step5: Stop the iteration. Feature Extraction In this method, from the segmented image diseased regions are selected and then apply Gray Level occurrence Matrix (GLCM). These matrices are used in classification process to retain the symptoms of diseases by using their feature vectors. Some of the Gray Level occurrence Matrix (GLCM) features are given in figure.3 Disease Category Figure.3 Feature Extraction Figure.2 Flow Diagram Image Pre-processing Images are taken from the farm fields and also the images are collected from the IRRI centre. The image pre-processing is performed from the collected input images. To minimize the noise and Figure.4 Disease category Detection and classification of Diseases Artificial neural network is used for detection and classification of diseases. The Artificial Neural Network (ANN) classifier includes three layers. They are input, hidden and output layers respectively. The number of inputs fed to the neural 3392

network gives the amount of texture features. The output is healthy, Leaf blast, brown spot and Bacterial leaf blight. 4.

This section clearly explains about the process and output in each stage.

affected region of leaf by enabling boundaries across the diseased leaf. Image Acquisition Image acquisition plays a vital role in capturing of input image from the external environment.

7 Boundary Detection Masking of disease affected portion is done with respect to four different stages as shown above.

3 network gives the amount of texture features. The output is healthy, Leaf blast, brown spot and Bacterial leaf blight. 4. Results And Discussions The following are the images obtained as results from Machine Vision system for disease recognition in paddy plant using basic features like color and texture. This section clearly explains about the process and output in each stage. Figure 7 interprets the masking based size feature namely masking the big particles of diseased leaf, the masking based on holes occurring on diseased leaf,masking based on erosion and the disease affected region of leaf by enabling boundaries across the diseased leaf. Image Acquisition Image acquisition plays a vital role in capturing of input image from the external environment. The input image acquired from external environment is fed as input to further stages and binaryequivalent of the image is obtained from basic RGB model. Figure.7 Boundary Detection Masking of disease affected portion is done with respect to four different stages as shown above. Several features are being taken into account for masking the disease affected portion of paddy leaf. Figure.5 Input Image A graph is plotted with respect to gray levels and pixel count of input image in x-axis and y- axis respectively. The histogram equalisation of all three bands namely red, green and blue are interpreted graphically with respect to gray level and pixel count parameters. Graphical Interpretation Fuzzy C-Means Clustering (FCMC) and frequency approximates of images are performed to identify and classify type of disease bothering the plant. Graphical interpretations of paddy leaf with respect to various parameters are performed to get a clear overview and analysis of stability and efficiency of different processes being operated. Correlation Graph Correlation is a major phenomenon in image processing that correlates input images trained in particular set of data. From the figure 8 it is observed that the image has high correlation with respect to the input image. Figure 6. Histogram Of Three Bands Boundary Detection To identify and diagnose the diseases in paddy leaf, the disease affected portions must be masked which helps in efficient classification of type of disease disturbing the leaf. This is made successful by boundary detection technique which masks the disease affected portion from healthy portion. Performance Plot Figure 8. Correlation Graph 3393

Performance plot for the trained, validated and test data is obtained which gives performance status of operating system. The figure 9 is evident for the mean square error performance. Figure.

This interprets the regression graph which portrays the error factor or deviation factor from actual Fama -French factor. Here we obtained the deviation factor is approximately R=0.

11 confusion Matrix Disease diagnosis The image is processed in various steps and divided into several parts based on physical quantities described earlier. This is achieved in predefined manner.

Experimental results depicts that 93.27% accuracy is obtained from this system. Confusion Matrix Figure.

4 Performance plot for the trained, validated and test data is obtained which gives performance status of operating system. The figure 9 is evident for the mean square error performance. Figure.9 Performance Plot Regression Plot There is always a standard deviation from output. This interprets the regression graph which portrays the error factor or deviation factor from actual Fama -French factor. Here we obtained the deviation factor is approximately R=0.753 which is clearly shown in figure 10. Figure.11 confusion Matrix Disease diagnosis The image is processed in various steps and divided into several parts based on physical quantities described earlier. This is achieved in predefined manner. The segmented image is analyzed with the help of machine vision or feature extraction method. The output of proposed system clearly shows the infected portions of a paddy. Experimental results depicts that 93.27% accuracy is obtained from this system. Confusion Matrix Figure.10 Regression Graph A 5x5 confusion matrix is plotted with respect to output classes of diseases and target classes. The row1 column1 of confusion matrix indicates the Healthy Leaf class. The row2 column2 of confusion matrix indicates the Leaf Blast disease.the row3 column3 of confusion matrix indicates the Brown Spot disease class. The row4 column4 of confusion matrix indicates the Bacterial Blight disease class. Figure.12 Result These diseased portions should be subjected to suitable treatment and diagnosis. Biological and chemical methodologies can be implemented provided by scientists and disease can effectively be cured. CONCLUSION The machine vision system for paddy leaf disease recognition is implemented using artificial neural network. The proposed method is suitable for detection of paddy leaf diseases. This method gives the disease detection accuracy of 93.27%.This method helps to farmers to make decisions in earlier stages and prevent from loss of production. 3394

5 REFERENCES [1] H. Syahputra,, A. Harjoko, R. Wardoyo and R.Pulungan, "Plant recognition using stereo leaf image using gray-level co-occurrence matrix", Journal of Computer Science 10, no. 4, (2013), p [2] P. Pallavi and V. S. Veena Devi Leaf regognition based on feature extraction and zernike moments, International journal of innovative research in computer and communication engineering, vol. 2, no. 2, (2014). [3] G. Ariputhiran and S. Gandhimathi, Feature Extraction and Classification of High Resolution Satellite Images using GLCM and Back Propagation Technique, International Journal Of Engineering And Computer Science, vol. 2, no. 2, (2013), pp [4] V. Metre and J. Ghorpade, An Overview of the Research on Texture Based Plant Leaf Classification, International Journal of Computer Science and Network, vol. 2, no. 3, (2013). [5] M. Maheshwari, M. Motwani and S. Silakari, New Feature Extraction Technique for Color Image Clustering, International Journal of Computer Science and Electronics Engineering, vol. 1, no. 1, (2013). [6] A.H. Kulkarni, H.M. Rai, K.A. Jahagirdar and P.S. Upparamani, A Leaf Recognition Technique for Plant Classification Using RBPNN and Zernike Moments, International Journal of Advanced Research in Computer and Communication Engineering, vol. 2, no. 1, (2013), pp [7] S. Naikwadi, N. Amoda, Advances in image processing for detection of plant diseases, International Journal of Application or Innovation in Engineering & Management, vol. 2, no. 11, (2013). [8] S. Abirami, V. Ramalingam, and S. Palanivel, Species Classification of Aquatic Plants Using PSVM and ANFIS, Pattern Recognition and Image Analysis, vol. 23, no. 2, (2013), pp [9] A. Kadir, L. E. Nugroho, A. Susanto and P. I. Santosa, Leaf Classification Using Shape, Color, and Texture Features, International Journal of Computer Trends and Technology, (2011), pp [10] Bhatt, Ashutosh Kumar, Durgesh Pant, and Richa Singh. "An analysis of the performance of Artificial Neural Network technique for apple classification", AI & society, vol. 29, (2014), pp [11] J. A. Jaleel, S. Salim, R.B. Aswin, Artificial Neural Network Based Detection of Skin Cancer, International Journal of Advanced Research in Electrical, Electronics and Instrumentation Engineering, vol. 1, no. 3, (2012). [12] N. Fadilah,, J. M. Saleh, H. Ibrahim, and Z. A. Halim, "Oil palm fresh fruit bunch ripeness classification using artificial neural network", In Intelligent and Advanced Systems (ICIAS), th International Conference, (2012). [13] R. Venkata Ramana Chary, D.Rajya Lakshmi and K.V.N Sunitha, Feature extraction method for color image similarity, Advanced Computing: An International Journal, vol.3, no.2, (2012). [14] H. Yu, J. Cao, W. Luo and Y. Liu, Image Retrieval of Wood Species by Color, Texture, and Spatial, International Conference on Information and Automation, (2009). [15] L. Zheng, J. Zhang, and Q. Wang, "Meanshift-based color segmentation of images containing green vegetation", Computers and Electronics in Agriculture, vol. 65, no. 1, (2009), pp [16] Pydipati, R., T. F. Burks and W. S. Lee, "Identification of citrus disease using color texture features and discriminant analysis", Computers and electronics in agriculture, vol.52, no. 1, (2006), pp [17] Fu, H., and Z. Chi. "Combined thresholding and neural network approach for vein pattern extraction from leaf images", IEEE Proceedings- Vision, Image and Signal Processing, vol. 153, no. 6, (2006), pp

6 3396

Plant Leaf Disease Detection using K means Segmentation

Volume 119 No. 15 2018, 3477-3483 ISSN: 1314-3395 (on-line version) url: http://www.acadpubl.eu/hub/ http://www.acadpubl.eu/hub/ Plant Leaf Disease Detection using K means Segmentation 1 T. Gayathri Devi,