Plant disease plays a crucial role in the reduction as well as degradation of production and yield in the area of precision agriculture and is a major concern for farmers and agriculturists. Hence, the detection and identification of diseases among the crops is essential. In this chapter, the CNN model for the identification and classification of different plant diseases through its leaf images is used. Four diseases such as ergot, downy mildew, blast, and rust in the pearl millet crops are considered in this work. The images of the pearl millet crop are considered for the five classes: healthy, ergot, downy mildew, rust, and blast. The dataset consists of 2074 images. The dataset is trained for the 30 epochs. The proposed approach is compared with the various existing methodologies such as naïve Bayesian, decision tree, support vector machine, and random forest. The simulation result shows that the proposed approach using the CNN outperforms the existing approaches in terms of accuracy and loss.
Top1. Introduction
The fulfillment of the dietary requirements of the human and animals is the main goal of the traditional farming system. Therefore, the farmers focus more on growing the healthy cereals like millet etc. instead of the high yield crops like wheat and rice. Considering the increasing trend of commercialization in the field of agriculture, the farmers are interested in producing the crops which are higher in terms of yield and are able to fulfil the dietary requirements. This increases the need to develop such a precision agriculture-based crop which is efficient in terms of financial gain as well as nutrition (Darwin, 2004; Jukanti et al., 2016; Le Mouël & Forslund, 2017).
Pearl millet is considered as the high yielding and nutritious crop. The quality and productivity of the pearl millet crop is negatively impacted by the diseases like rust, blast, downy mildew, and ergot. The applicability of Internet of Things (IOT) in the collection and storage of data, and processing and integration of machine learning methods in the areas of object detection, recognition and visualization and pattern identification has motivated the authors to develop a machine learning based solution for identifying and categorizing plant diseases from the leaf images. The main contribution of the chapter is as follows:
- 1.
The chapter presents an automated approach to detect and classify the pearl millet disease from its leaf images.
- 2.
For the simulation purpose, the data augmentation technique is applied for the acquisition of image dataset.
- 3.
The different machine learning based approaches were applied and evaluated for the identification and categorization of the diseases in the pearl millet leaf images.
- 4.
The various approaches are compared for the plant disease identification and categorization.
The chapter is structured as follows: discussion of the work done in the field in the section 2, propose methodology is described in section 3, Section 4 includes a discussion of the implementation details and results, and Section 5 wraps up the study with a look into the future scope.
Top2. Literature Review
The literature study in the field of plant disease classification and detection provides the significant insight in the object detection, identification and visualization. The traditional approaches utilized the manual identification and categorization of the plant diseases. But the traditional approaches did not give the sufficient insights of different parameters of the crop such as temperature, soil moisture, humidity etc. and the plant actual status monitoring is difficult. In addition to this, the legacy approaches are usually time taking and requires extensive human effort. The farmers also required the expert advice for the correct detection of the crop health. The integration of IOT with the machine learning techniques have enabled the farmers to automatically detect the plant disease by observing the leaf images. The different IOT based solutions such as smart phones, GPS, drone cameras were used for the data collection, storage and processing. The work done in the field of plant disease detection are summarized as follows:
In Kitpo and Inoue (2018), authors have used the IOT based architecture for data collection. The authors have implemented the SVM based model for the detection of plant diseases in the rice plant. In Thorat et al. (2017), authors have used the IOT based solution for the classification of plant leaf images as healthy or unhealthy. In Lu et al. (2017), authors have proposed the CNN based approach for the detection and classification of 10 classes of disease in the rice crop. The simulations were carried out on the dataset of 500 images and the results show that the CNN based approaches achieves the higher accuracy over the traditional machine learning approaches. In Amara et al. (2017), authors have implemented the deep learning-based model Lenet to automatically detect sigatoka and speckle diseases in banana leaf. The proposed approach achieves the accuracy of 98% for the color images and 94% for the gray scale images. In Ramcharan et al. (2017), authors have proposed the deep learning approaches for the identification of diseases in three classes and two classes of pest damages on the casava image dataset.