Any developing economy can be thought of as having its foundation in the agricultural sector. One industry where automation and the internet of things can have a significant impact is agriculture and modern farming. Due to the country's expanding population and rising food demands, agriculture is important in India. As a result, agricultural yield needs to be increased. Diseases brought on by bacteria, fungus, and viruses are one of these significant factors in decreased crop yields. By using methods for plant disease detection, this can be avoided and managed. Farmers must have access to the latest technologies and practices in order to get the most yield possible from their crops. A maximum agricultural output must be maintained, which means keeping healthy plants and keeping an eye on their surroundings to spot or detect diseases. Artificial intelligence (AI), machine learning, and deep learning have all been proven to be incredibly useful in the field of modern agriculture as a means of sophisticated image analysis.
TopLiterature
D.A. Bashish et al. (2010) divided the leaf picture into four clusters using k-means segmentation and squared Euclidean distances. The Colour Co-occurrence approach was used to feature extraction for both colour and texture characteristics. Using a neural network detection method built on the Back Propagation methodology, classification is finally completed. The whole system's accuracy in classifying and detecting illnesses was estimated to be 93%.
M. Banghe and associates (2015) A web-based tool for diagnosing fruit ailments has been created as a result of user contributions of fruit photos to the system. The characteristics were extracted using CCV (colour coherence vector) metrics, morphology, and colour. The clustering process employed the k-means method. SVM is employed to ascertain a person's infection status. The accuracy of this study's diagnosis of pomegranate disease is 82%.
To detect fungal infections on plant leaves, J.D. Pujari et al. (2015) employed a range of crop types, including commercial crops, cereal crops, and fruit and vegetable crops. There have been several crop management tactics employed. • K-means clustering was used to segment fruit crops, and texture features were focused on and identified using ANN and closest neighbour algorithms, attaining an overall average accuracy of 90.723%.
- •
The chan-vase strategy is used for segmenting vegetable crops, local binary patterns are used to extract texture features, and SVM and the k-nearest neighbour algorithm are used to extract texture features. Overall average accuracy of 87.825% in classification
- •
Commercial crops were segmented using the grab-cut algorithm. With an overall average accuracy of 84.825%, wavelet-based feature extraction was used with Mahalnobis distance and PNN as classifiers.
- •
K-means clustering and the canny edge detector were used to segment the cereal crops. Colour, shape, texture, texture colour, and random transform features were retrieved. SVM and closest neighbour classifiers were applied, yielding a total accuracy of 83.72%.
In order to automate the detection and categorization of plant diseases, V. Singh et al. (2016) employed an image segmentation approach called a genetic algorithm. Using a limited number of images, four plant leaves—banana, rose, and lemon—were taught and evaluated.. beans, The colour co-occurrence strategy, which considered both texture and colour data, was used for feature extraction. The diseases were classified with 86.54% and 95.71% accuracy using the Minimum Distance Criterion with k-mean clustering and the SVM classifier, respectively. The accuracy increases to 93.63% when the genetic algorithm is paired with the Minimum Distance Criterion classifier.
E. Kiani et al. (2017) employed a fuzzy decision maker to spot diseased leaves in an outdoor strawberry crop. The detection time was 1.2 seconds, and the total accuracy of detection and segmentation of plant illnesses was 97%.