DETECTION OF PLANT LEAF DISEASES IN AGRICULTURE USING RECENT IMAGE PROCESSING TECHNIQUES – A REVIEW

Purpose: Agricultural productivity is something on which the economy highly depends in India as well in all over the world. India is an agriculture-dependent country; wherein about 70% of the population depends on agriculture. Methodology: This is one of the main reasons that disease detection in agriculture plays an important role, as having the disease in plant leaf is quite natural. If proper observations are not taken in the agriculture field then it causes serious effects on plants due to which respective product quality and productivity are affected. Detection of plant leaf disease through effective and accurate automatic technique is beneficial at the starting stage as it reduces a large work of monitoring in big farms of crops. Result: This paper presents the review on the state of the art disease classification techniques presently used using image processing that can be used for plant leaf disease detection in agriculture.


INTRODUCTION
India is an agricultural country; agriculture has a very large contribution to the economy of this country. The disease in the leaves of the plants is a matter of great concern for any farmer as well as for the higher scale farming. Knowing disease at the beginning stage is a very difficult task with naked eyes, it is very important for someone to know this work very well to identify which type of disease started, because of which the crops sometimes get infected because they are not treated at the right time. This research paper shows the research work done to solve this problem in the past years and tried to understand how the methodology of image processing detects the disease at an early stage.

LITERATURE REVIEW
The analysis of the research works done for the consolidation of this problem is summarized as follows in table 1.  Truong et.al (2017) exhibited the plan about analyzing the fungal disease on crops due to environmental effect by using real-time approach, internet of things and prevent it by using machine learning in which SVM (support vector machine regression) algorithm is used to process the data and predict day by day according to the environmental parameter. This paper shows the design of an Internet of Things (IoT) system consisting of equipment that is capable of sending realtime environmental data to cloud storage and a machine-learning algorithm to predict environmental conditions for fungal detection and prevention. Here a rulebased semi-automatic system using concepts of k-means is designed and implemented to distinguish healthy leaves from diseased leaves. A diseased leaf is classified into one of the three categories (downy mildew, frog eye, and Septoria leaf blight). Testing operations are performed by separately utilizing color features, texture features, and their combinations to train three models based on the support vector machine classifier. Results are generated using thousands of images collected from the Plant Village dataset. This study also trying to discover the best performing feature set for leaf disease detection in Soybean.
Singh, G., Kumar, G., Bhatnagar, V., Srivastava, A., & Jyoti, K. (2019) developed an environment to control and detect the environmental pollution using IoT. It includes different sensors that work and sense data and submit the data to the main server.

DETECTION METHODOLOGIES
Various leaf diseases cause severe losses to farmers resulting in a major threat to the growers. To minimize losses, a support system is required to detect plant leaf diseases. The image processing approach is a non-invasive technique that provides a reliable, cost-effective, and accurate solution to the farmers in minimal time to optimize the yield losses. Basic steps for the identification and classification of plant disease detection and classification are shown in the following figure 1.
 Input Image set: Plant leaf images will be imported from the field directly or from the different database like plant leaf village (Sharda P. Mohanty, David P. Huges and Marcel Salathe (2016)) and (Konstantios P Ferentinos (2018)) available from the research center and institution of the government firms.  Image pre-processing: This step is used to specify a suitable color transformation that best highlighted the diseased regions shown in the picture using different tools available in the MATLAB.
 Image Enhancement: This step is used to develop the filter that could highlight those regions considered target (target diseased area).
 Image Segmentation: This step is to recognize the region that was likely to qualify as the diseased region in the image. For the segmentation of the target part, so many classifier algorithms are present in which k-means clustering (G Li, Z Ma H Wang, (2016)), (Çaglar Küçük, Gül¸ sen Ta¸ skın, and Esra Erten (2016)), (A Sabu, K Sreekumar (2017)), (Sukhvir Kaur, Shreelekha Pandey, and Shivani Goel (2018)) is widely used algorithms.
 k-Means Clustering Algorithm  (2018)) procedure, which is as follows: 1. Consider k of the r given patterns to each form of clusters. We thus have k clusters, each with one pattern. The remaining patterns remain as such.
2. The k patterns to form clusters may be chosen arbitrarily: they are usually the first k patterns. In this step, the centroid of a given cluster is the coordinates of the pattern in the cluster in the A1-A2-…….-AM coordinate space.
3. For i = k+1, k+2 ,……, r do. Put the pattern in the cluster whose centroid is the nearest to the pattern. Compute the cluster's new centroid.
Step 1: Let C' be the cluster that has the i th pattern. Calculate the distance between the i th+1 and the centroids of each of the k Clusters. If the pattern is closet to the centroid of C', then pattern need not changed its cluster, hence go to step 3.
Step 2: Let C" be the cluster whose centroid is closest to the i th pattern. Move the pattern from cluster C' to C". Compute the new centroids of clusters C' and C" . Go to Step 4.
Step 4: No patterns changed clusters in the last iteration; hence return from the procedure with the k clusters.
 As should be clear from the above steps, the k-means procedure begins by putting each pattern in one of the k clusters (steps 1 and 3). Then, in step 4, it ensures that every pattern is in a cluster is closest to it.
 The procedure is said to be the partitional because it partitions the set of clusters into clusters.
 Feature Extraction: After Segmentation of the region of interest selected which is having better image data using various features extracted from different feature extraction techniques. Reducing the amount of resources required to describe a large dataset. Here we extract color features and shape features. For example, the target class may be a variety of diseases, such as mildew, late blight and early blight, which are classified based on their texture or color features.

CONCLUSION
This paper focused on plant leaf disease detection and classification methods using image processing, machine learning, and deep learning. Through this review, concluded that the plant disease detection techniques consist of common three steps which are pre-processing, segmentation and feature extraction, and classification. Image segmentation is performed by the K-means clustering algorithm. Deep learning models are also introduced in recent times for the detection of the plant leaf disease and found that convolutional neural network through deep learning classifier gives the 99.53%