Plant Disease Classification with Neural Networks

ABSTRACT

The plant disease diagnosis is restricted by person’s visual capabilities as it is microscopic in nature. Due to optical nature of plant monitoring task, computer visualization methods are adopted in plant disease recognition. The aim is to detect the symptoms of the disease occurred in leaves in an accurate way. Once the captured image is pre-processed, the various properties of the plant leaf such as intensity, colour and size are extracted and sent to classifier with Artificial Neural Network for classification of disease which the plant gets.

INTRODUCTION

The Production of good quality food produce and improvement in crop yield are challenging for the researchers as well as agriculturist to meet the growing demands globally. Thus, it is crucial to maximize agriculture resources and its utilizations in a sustainable manner. Therefore, for the sustainable agriculture system, use of emerging technology becomes important for significant and efficient contributions. With the implementation of these techniques, possibility to reducing errors and costs for achieving ecologically and economically sustainable agriculture is the thought of the present era [1]. Earlier used techniques were inefficient and time consuming for analyzing the problems and implementation of remedial measures. Diseased plants exhibit a variety of symptoms like, stunting, yellowing, wilting, twisting, reddening, browning, blighting, and other abnormalities [2]. Thus, accurate diagnosis is essential to diagnosis and control the plant disease effectively. Until a disease is adequately diagnosed, a grower may waste time and energy as well as money to solve a problem with an unknown cause. Once a disease is diagnosed, appropriate management practices can be selected [3]. To overcome this problem a fast and accurate process is required, that can automatically detect the disease on the leaf. Technique such as visual detection requires significant time for visual inspection for a large cultivated area. Thus, image processing technique is proven to be an effective method as compared to visual analysis.

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 an order

LITERATURE SURVEY

In the modern era, enhancement in the use of internet attracts the science and engineering techniques to get easy and quick solution, as, it is most efficient and effective way of communication. Therefore, researcher, Bhange et al. (2015) developed a web based tool for identifying pomegranate leaf disease. In the first step, feature extraction (based on color and morphological features) was done. Thereafter, to segment the diseased part from healthy region k-mean algorithm and SVM was used. Accuracy achievement in the proposed method was 82% [4]. Renugambal et al. (2015) proposed an artificial intelligence technique for automatic detection and classification of Sugarcane leaf disease using image processing technique. Infected leaves were captured by using digital camera. After then, the preprocessing and segmentation was done using image histogram equalization, filtering, color transformation to detect infected parts of the leaves. Finally, SVM classifier was used for classification purpose [5].

An image segmentation algorithm was proposed by Singh and Misra (2017) for automatic detection and classification of plant leaf disease (rose, banana, beans). Segmentation was done using genetic algorithm to distinguish diseased part and the healthy parts of the leaves. This algorithm was tested on ten species of plant i.e. Jackfruit, Banana, Mango, Sapota, Potato, Beans, Tomato, Lemon, etc. to check the accuracy of proposed algorithm. The author accomplished that the algorithm provides optimum results with less computational efforts in recognition and classification of leaf disease [6].

PROPOSED METHOD

IMAGE ACQUISITION

The impression of images is the primary and essential step to observe the state of the Groundnut leaf. The image imprisonment has been done through various tools and devices, such as, cameras, mobile phones and satellites. The proper estimation of RGB color pixels in an image is essential step towards successful completion of image capturing. The technical parameters of these simple, handheld devices such as light sensitivity of the photo sensors, spatial resolution and digital focusing have improved dramatically year after year. Today, nearly every person, farmer or plant pathologists carry these modern and sophisticated devices such as digital cameras together with a mobile phone or tablet computer.

IMAGE PRE-PROCESSING

The pre-processing follows the image acquisition. The acquisition of images and creating images database, pre-processing has been done. The pre-processing of created database is a preliminary step to eliminate the undesired distortion of the image and provides enhancement in features. While considering leaf of a plant, various colours have been observed. To distinguish the colour of the diseased lesion from the original colour of leaf, the RGB colour pixels should be converted into some other pre-processing for the better perception. The reason for unacceptability of RGB is the system dependency of such pre-processing. Therefore, the improvement in the precision of colour for detection of disease, the independency of pre-processing is essentially required.

IMAGE SEGMENTATION

Segmentation of an image is the process of partitioning the object (diseased spot) from its background (leaf). Different segmentation techniques are available like clustering methods, thresholding, edge detection, ANN based methods, partial

differential equation based segmentation, etc. In the present research kmean clustering technique for segmentation has been given the priority among all of the above stated techniques. The inherent advantages of k-mean clustering method are that, it works well with large data sets. The accuracy of system depends on the data sets. Therefore, this (k-mean clustering) proves to be fast, robust, easier to understand and simplest to implement. Furthermore, it may work more efficiently; if clusters are spherical (diseased spots are spherical in shape) and more in number. Increased value of the k (cluster) reduces the amount of error in the result. Current work has a value of five for k.

The formations of clusters have been done based on the selection of five random points selected from the data sets. These five random points treated as centroids of each cluster. These random points attract the same intensity points (based on Euclidian distance method). This movement of the centroid happened till the same intensity cluster formed and can’t move further. The ultimate end results come in the form of diseased and healthy parts of the leaves. After segmentation, one of the diseased clusters (obtained from one or more than one cluster) has been extracted and considered for calculation of the disease area of the leaf.

FEATURE EXTRACTION

At this stage of the project we calculate the Gray Level Co-occurrence Matrix of an image in order to extract the set of features required for further calculations. In a statistical texture analysis, texture features were computed on the basis of statistical distribution of pixel intensity at a given position relative to others in a matrix of pixel representing image. Depending on the number of pixels or dots in each combination, we have the first-order statistics, second-order statistics or higher-order statistics. Feature extraction based on grey-level co-occurrence matrix (GLCM) is the second-order statistics that can be used to analyze the image as a texture. GLCM (also called gray tone spatial dependency matrix) is a tabulation of the frequencies or how often a combination of pixel brightness values in an image occurs. Transforming the input data into the set of features is called feature extraction. In this project, the features are Mean, SD (Standard Deviation), Entropy, RMS, Variance, Smoothness, Kurtosis, Skewness, IDM, Contrast, Correlation, Energy, Homogeneity. If the features extracted are carefully chosen it is expected that the features set will extract the relevant information from the input data in order to perform the desired task using this reduced representation instead of the full size input.

Artificial Neural Network (ANN)

The key element of this paradigm is the novel structure of the information processing system. It is composed of a large number of highly interconnected processing elements (neurones) working in unison to solve specific problems. ANNs, like people, learn by example. An ANN is configured for a specific application, such as pattern recognition or data classification, through a learning process. Learning in biological systems involves adjustments to the synaptic connections that exist between the neurones. This is true of ANNs as well. The below fig-2 shows the diagram of ANN(Artificial Neural Networking).

CONCLUSION

There are many methods in automated or computer vision plant disease detection and classification process, but still, this research field is lacking. In addition, there are still no commercial solutions on the market, except those dealing with plant species recognition based on the leaves images. By observing the existing and proposing method, some differences are shown in the table-1.In this paper, a new approach of using deep learning method was explored in order to automatically classify and detect plant diseases from leaf images. The developed model was able to detect leaf presence and distinguish between healthy leaves and 4 different diseases, which can be visually diagnosed. The complete procedure was described, respectively, from collecting the images used for training and validation to image preprocessing and augmentation and finally the procedure of training the deep ANN and fine-tuning. Different tests were performed in order to check the performance of newly created model.

References

1. K. A. Garrett, S. P. Dendy, E. E. Frank, M. N. Rouse, and S. E. Travers, “Climate change effects on plant disease: genomes to ecosystems,” Annual Review of Phytopathology, vol. 44, pp. 489– 509, 2006.
2. S. M. Coakley, H. Scherm, and S. Chakraborty, “Climate change and plant disease management,” Annual Review of Phytopathology, vol. 37, no. 1, pp. 399–426, 1999.
3. S. Chakraborty, A. V. Tiedemann, and P. S. Teng, “Climate change: potential impact on plant diseases,” Environmental Pollution, vol. 108, no. 3, pp. 317–326, 2000.
4. A. J. Tatem, D. J. Rogers, and S. I. Hay, “Global transport networks and infectious disease spread,” Advances in Parasitology, vol. 62, pp. 293–343, 2006.
5. J. R. Rohr, T. R. Raffel, J. M. Romansic, H. McCallum, and P. J. Hudson, “Evaluating the links between climate, disease spread, and amphibian declines,” Proceedings of the National Academy of Sciences of the United States of America, vol. 105, no. 45, pp. 17436–17441, 2008.
6. T. Van der Zwet, “Present worldwide distribution of fire blight,” in Proceedings of the 9th International Workshop on Fire Blight, vol. 590, Napier, New Zealand, October 2001.
7. . A. Miller, F. D. Beed, and C. L. Harmon, “Plant disease diagnostic capabilities and networks,” Annual Review of Phytopathology, vol. 47, pp. 15–38, 2009.
8. M. B. Riley, M. R. Williamson, and O. Maloy, “Plant disease diagnosis. The Plant Health Instructor,” 2002.
9. J. G. Arnal Barbedo, “Digital image processing techniques for detecting, quantifying and classifying plant diseases,” SpringerPlus, vol. 2, article 660, pp. 1–12, 2013.
10. H. Cartwright, Ed., Artificial Neural Networks, Humana Press, 2015.