An Automated Detection Of Visual Plant Disease Through Texture Analysis

Abstract

Pakistan is an agricultural country and the economy of Pakistan is highly depending on its agriculture productivity. Detection of diseases in plants plays a significant role in the agriculture field of Pakistan. Lack of timely and proper care of plants massively affects the output quality, quantity, and productivity. This paper presents an algorithm for the detection and identification of plant leaf disease on the basis of image segmentation and texture analysis. In the first stage, image segmentation is performed to segment the leaf from the background. Afterward, feature extraction is performed on a preprocessed leaf image. We extract local and global texture features. Support vector machines classification is applied on extracted features to perform discrimination between healthy and diseased leaf classes. Results of our proposed framework show promising results which outperform the existing work in terms of accuracy and computational cost.

Introduction

Plant diseases have serious effects on the quality and quantity of plants by destroying their normal state. These diseases interrupt plant’s vital functions such as photosynthesis, transpiration, pollination, fertilization, and germination. Correct recognition and diagnosis of plant diseases at the very first stage is very important. The symptoms of plant diseases normally appear on plant’s leaves; thus it is possible to automatically detect crop diseases by applying extracting features, image segmentation and machine learning techniques on leaf images. Specifically, plant diseases can be classified and recognized by analyzing color, texture features, and shape of the diseased leaf images.

Most existing image-based plant leaf disease recognition algorithms based on extracting various kinds of features from leaf images of affected plants. These algorithms have common restrictions as the features selected for discriminating leaf images are usually treated as quality important in the classification process. We are working on plants disease recognition which will be comprised of four major procedures that are pre-processing of acquired images using Camera, remove noise from image and perform contrast enhancement on diseased plant leaf images, extracting features containing information about disease using Local Binary Pattern (LBP), segments and classifying diseased leaf images using Support Vector Machine (SVM

LITERATURE REVIEW

In this paper, authors were described their work on different classification techniques that are used for plant leaf disease recognition & classification. Authors used, K-nearest neighbor method for this purpose because it is suitable as well as the simplest among all algorithms for class calculation. Drawbacks of their work are that SVM not show optimal parameters if training data is not linearly divisible on it. [1].

Authors have used four processing steps for detection of diseases from plant leaf image using image processing. First is for the input RGB image, a color transformation structure is formed. HIS is used for the color descriptor. In the 2nd step, by using threshold value, green pixels are masked and removed. In 3rd step by using pre-computed threshold level, removing of green pixels and masking is done for the useful segments that are removed first in this step, while the image is segmented. In the last step, the segmentation is done [2].

In paper [3], authors were used neural networks technique for automatic detection of disease in plants leaves. This proposed approach significantly performed accurate detection of disease in leaf, Stem and root diseases required very efforts in their computation that is its drawback.

In Paper [4], authors were proposed their work on disease identification in four processing steps. In the first step, RGB image and color transformation structure is taken and using a specific threshold value, the green pixels are masked and extracted from the image, which is further followed by segmentation process, and for getting useful segments the texture statistics are calculated. At last, a classifier is used for the features that are removed to classify the disease. The strength of the proposed algorithm is proved by using the experimental result of about 500 plant leaves in a database.

In this paper authors were used a methodology for early and accurate plant leaf disease detection, Researchers worked on artificial neural network (ANN) and various image processing techniques for this purpose. Authors were used ANN classifier that classifies various plant disease using a combination of feature, texture, and color of the image.[5].

Save your time!
We can take care of your essay

• Proper editing and formatting
• Free revision, title page, and bibliography
• Flexible prices and money-back guarantee

Place Order

In this paper K-mean clustering, texture and color analysis method were used by the authors for the detection of disease from leaf image[6].

In paper [7] author used histogram matching technique for finding disease from plants leaves. The author says in plants diseases seems on leaves first, therefore the histogram matching technique is done according to an edge detection method and color features.

In the study of paper [8], the authors proposed two methods that are triangle threshold and simple threshold methods which are used to separate the lesion region area and leaf area individually. In the final step, categorization of disease is done by calculating the measure of leaf area and lesion area. According to the research done the proposed method is fast and accurate for finding leaf disease.

Researchers presented an algorithm for disease spot detection in plant leaf using image processing techniques. In this paper process of disease, spot detection is done by comparing the effect of HIS, CIELAB, and YCbCr color space. For image soothing Median filter is used. At their final step, they applied the Otsu method on a color component, calculation of threshold to find the diseased spot. There is some noise because of background which is shown in the experimental result, camera flash and vein. CIELAB color model is used to remove that noise. In paper [9] authors used features extracted from leaf images, many crop diseases recognition methods and systems have been developed.

According to paper [10], the state of art review of different methods for leaf disease segmentation used image processing techniques. In the paper [11] authors in their research used different leaf image processing and recognition technologies to study the cucumber disease of downy mildew, powdery mildew, and anthracnose. Authors in their paper used auto-cropping segmentation and fuzzy contents classification to analysis leaf color in the olive leaf spot disease [12].

Proposed Methodology

Different types of image processing, feature extraction and classification techniques are used to process those images taken by the digital camera to obtain different and useful features needed for the analyzing of disease. The procedure for the proposed image processing and segmentation algorithm is described below.

1. Taking the image of plant leaf with the help of a digital camera.
2. In the preprocessing, remove noise, contrast enhancement, color enhancement remove distortion techniques are used for improving the quality of an image.
3. Apply the texture features i.e. Local Binary Pattern (LBP) for extraction of useful information contained by the image
4. Classification is the final step, in which Support Vector Machine (SVM), K-Nearest Neighbor (KNN) methods are used to classify the image.

Feature Extraction using Local Binary Patterns

The basic idea of Local Binary Pattern abbreviated (LBP) is to summarize the local structure in the image by comparing each pixel with its neighborhood. Take a pixel as center and threshold against its neighbors. If the intensity of the center pixel is greater than or equal to its neighborhood, then denotes it’s with 1 and place 0 if not. You will receive with a binary number for each pixel, just like 11100010. You will end up with 2^8 possible combinations with eight surrounding pixels.

Conclusion

In this paper, we proposed a method for automated detection of plant leaf using LBP and apply SVM for classification. We proposed method achieved higher overall accuracy as compared to the existing research. We compared the performance of different classification methods. There are limitations to our work, but In the future, we aim to apply enhanced feature extraction methods to achieve better accuracy.

References

1. S. N. Ghaiwat, P. J. I. J. o. R. A. i. E. Arora, and Technology, 'Detection and classification of plant leaf diseases using image processing techniques: a review,' vol. 2, no. 3, pp. 1-7, 2014.
2. S. B. Dhaygude, N. P. J. I. J. o. A. R. i. E. Kumbhar, Electronics, and I. Engineering, 'Agricultural plant leaf disease detection using image processing,' vol. 2, no. 1, pp. 599-602, 2013.
3. M. R. Badnakhe and P. R. Deshmukh, 'An application of K-means clustering and artificial intelligence in pattern recognition for crop diseases,' in International Conference on Advancements in Information Technology, 2011.
4. S. Arivazhagan, R. N. Shebiah, S. Ananthi, and S. V. J. A. E. I. C. J. Varthini, 'Detection of unhealthy region of plant leaves and classification of plant leaf diseases using texture features,' vol. 15, no. 1, pp. 211-217, 2013.
5. A. H. Kulkarni and A. J. I. J. o. M. E. R. Patil, 'Applying image processing technique to detect plant diseases,' vol. 2, no. 5, pp. 3661-3664, 2012.
6. S. Bashir, N. J. I. J. o. E. Sharma, and C. Engineering, 'Remote area plant disease detection using image processing,' vol. 1, no. 6, pp. 31-34, 2012.
7. S. Naikwadi, N. J. I. j. o. a. o. i. i. e. Amoda, and management, 'Advances in image processing for detection of plant diseases,' vol. 2, no. 11, 2013.
8. S. B. Patil, S. K. J. I. J. o. E. Bodhe, and Technology, 'Leaf disease severity measurement using image processing,' vol. 3, no. 5, pp. 297-301, 2011.
9. P. Chaudhary, A. K. Chaudhari, A. Cheeran, S. J. I. J. o. C. S. Godara, and Telecommunications, 'Color transform based approach for disease spot detection on plant leaf,' vol. 3, no. 6, pp. 65-70, 2012.
10. A. N. Rathod, B. Tanawal, V. J. I. J. o. A. R. i. C. S. Shah, and S. Engineering, 'Image processing techniques for detection of leaf disease,' vol. 3, no. 11, 2013.
11. D. Pixia and W. Xiangdong, 'Recognition of Greenhouse Cucumber Disease Based on Image Processing Technology %J Open Journal of Applied Sciences,' vol. Vol.03No.01, p. 5, 2013, Art. no. 26952.
12. M. S. J. W. A. S. J. Al-Tarawneh, 'An empirical investigation of olive leave spot disease using auto-cropping segmentation and fuzzy C-means classification,' vol. 23, no. 9, pp. 1207-1211, 2013.