Image Processing Algorithms For Leaf Disease Detection Classification and Severity Measurement
Loading...
Date
item.page.authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The agricultural dependency of economy of India shows the significance and
newlinerequirements in advancement of farming methodologies and their respective
newlineimprovisation leading to impactful economic advancement. Corps disease as main
newlineconstraint and natural cause of productivity degradation triggers the thought of
newlinefarming technology advancement to get priority based necessary treatments and action
newlineplans. The sole responsibility lies on the detection and identification of plant diseases
newlinefor necessary action planning. The combination of neural network and image
newlineprocessing methods implemented using infected leaf images for disease detection and
newlineclassification. Further this image analysis can provide sufficiently early disease
newlineinformation to follow the necessary actions needed that can avoid the further
newlinedisastrous situations and economic losses.
newlineIn this research, auto-segmentation using Fast Adaptive Fuzzy C-means
newlineClustering (FAFCM) and Incremental K-means clustering methods employed for leaf
newlineimage segmentation. The combined LBP and HOG feature technique employed for
newlinerobust and accurate identification of Leaf species. The Incremental K-means
newlineclustering method with RGB color and GLCM texture feature employed using
newlineshallow classifiers for Soybean, Tomato, and Potato, leaf species disease detection,
newlineand classification.
newlineThe Deep Convolutional Neural Networks (CNN s) techniques like AlexNet,
newlineGoogleNet, VGG16, ResNet101, and DensNet201 are used to increase the
newlineperformance of disease detection and classification system. For method validation,
newlineself-developed Soybean leaf image data set used with Plant village image public
newlinestandard dataset. The experimental result shows that the Combination of GoogleNet
newlineand VGG16 outperforms with 96.4% accuracy each than others. Compared with
newlineshallow classifier proposed pre-trained CNN s would improve accuracy and
newlinerobustness in leaf species disease detection and classification.
newlineAlso, the Combined LBP and HOG (LBHPG) features that would improve
newlineaccuracy and robustness in l