Article Preview
Top1. Introduction
Plants and fruits are primary energy sources for humans and animals. The plant leaves play an essential role in the growth of plants through photosynthesis. The leaves of the plants are also beneficial for humanity because of their medicinal properties (Dhingra et. at 2018). In the Asian and African countries, more than 50% population directly depends upon agriculture for food, shelter, medicine, and employment (Patel et al., 2017). The economy and living of the small farm holders entirely depends upon the production of agriculture products. However, any type of disease degrades the quality of crops and agriculture products and hence leads to huge economical loss of framers. Various crops are damaged because of diseases which declines the quantity and quality of agricultural production. Plant diseases are broadly categorized into parasitic diseases and non-parasitic diseases. The parasitic diseases may occur due to various pathogens such as viruses, bacteria, fungus, chromista, etc.; pests such as millet, mammals, slugs, rodents, etc; weeds such as monocots and dicots. In contrast, non-parasitic plant diseases may occur due to excess or shortage of water, temperature, irradiation, minerals, and nutrients (Kaur et al., 2019) ( Poojary et.al 2019).
Crop diseases are chief hazard to food security, but disease detection is challenging in many parts of the world as a result of the unavailability of infrastructure. Recent growth in smartphone technologies paved the way for computer vision-based disease detection. Plant diseases establish financial disasters to the smallholder farmers whose livelihood depends on vigorous crops (Harvey et al., 2015).
The manual leaf disease detection technique is time-consuming, tedious, and depends upon expert knowledge. The manual leaf disease detection is often subjected to less accuracy and inefficient because of fatigues, tiredness, and lack of interpretation of disease. Thus, image processing-based automatic machine learning and deep learning methods are used for plant leaf disease detection techniques (Chapaneri et al., 2020)( Joshi et.al 2020)(Kaur et al 2019).
Machine learning models can be used for thing identification, review, prediction, categorization, and grouping of object. The learning task in ML categorized as “Supervised”, “unsupervised” and “semi-supervised”. Depending upon the problem statement different models is used.
Deep neural networks are advances of the neural network that have been successfully applied recently for many computer vision-based applications. Deep neural networks are constructed by stacking the series of layers of nodes. The deep learning algorithms' performance can be improved by tuning the parameters of the deep learning layers. The accuracy of deep learning models depends on the size of the database (LeCun et al., 2015)( Schmidhuber et al., 2015).
This paper presents multi-class plant leaf disease detection based on a deep convolutional neural network (DCNN). Three-layered DCNN architecture that consists of convolution, rectified linear unit layer, maximum pooling layer (MP), fully connected layer, and a classification layer is used. The database contains healthy and disease samples of tomato, potato, and pepper plant leave . Various diseases consider for the implementation are bacterial spot, early blight, late blight, target spot, mites, leaf mold, and septoria leaf spot.
The objectives of this article are to are as follows:
- •
To design and implement simple three layered Cascaded Deep Convolutional Neural Network (CDCNN) model.
- •
To improve the feature distinctiveness of the leaf image.
- •
Effectively detection of diseases not only for a particular plant but for multiclass plant.
- •
Testing the designed model in the real itme surroundings.