In this method, the details of the appropriate shape information are expressed in a pattern so that the task of arranging the pattern is made easy. It is used in pattern recognition and a specific form of size reduction [5]. The process of feature extraction is helpful when you need to lessen the number of stratagems needed for processing without losing relevant information. Feature extraction can also depict the amount of surplus data for a given analysis.
Classifiers
Classifiers are used to identify the presence of disease-affected region in a leaf. These classifiers use few methods for detection of disease. There are three methods used as classifiers:
- Percentage infection
- Support vector machine (SVM)
- Back-propagation neural network
Existing system
Conventionally, farmers can determine changes in the leaf color to point out the diseases by the naked eye inspection method. Generally, farmers use the naked eye to detect the diseases in plants by observation method. The detection of plant diseases is done by the trained farmer, who can detect even the small changes in the color of the leaf. The observation method to detect the disease in the paddy is time-consuming, laborious and not possible in large area agriculture fields. Best methods and practices need to be implemented for detecting plant leaf diseases, which are accurate, fast and helpful to the farmers.
Widespread diseases in paddy leaf are narrow brown-spot disease, brown-spot disease and blast disease (BD). To determine these diseases, we have various techniques. In the past decade, research using image processing was conducted to detect and analyze plant diseases [6]. Image processing techniques are aimed to decrease the subjectiveness and enhance the throughput to detect paddy diseases. The detection of the diseased part of the leaf is basically done in two steps. The first step is image acquisition and the second step is segmentation using spot detection method and boundary detection. The classification of diseased paddy leaf is based on zoom algorithm. The zoom algorithm works on self-organizing map neural network.
We have two processes to build self-organizing map of the input vector. The first process uses the zeros to padding, and the second process uses the detection of missing points in interpolation. The interpolation method is used to normalize the spots by zoom algorithm. To apply the zoom algorithm for the segmentation, the first step is image acquisition and the second one is the K-means clustering method. Color co-occurrence method is applied to analyze the infected leaf and stem. Finally, the algorithm of back propagation is used with the neural network for classification diseased plants [8]. Using image processing methods, the identification and categorization of plant diseases is done with accuracy of around 94.00%.
Some other techniques for determination of the disease constitute image acquisition, color processing, feature extraction and classification by considering the production rule technique with a forward chaining method. Precision is around 92.00% for this method, and it is not a recommended technique for disease identification.
Suggested method
In this method, the identification of plant disease is done by internet of things–based technique. In this technique, sensors are placed at different places across the field, these sensors collect data and send it to a processing unit where images are processed with different methods to identify the disease-affected plants by examining the color of the leaf. This process is done by feature extraction and machine learning techniques. The suggested method is color image processing, along with three different techniques: (i) image segmentation, (ii) image smoothing, (iii) feature extraction. We got 96% accuracy in disease classification by using the vector machine demonstration. Our approach is towards massive classification of plant diseases with automated techniques.
Leave a Reply