Abstract
Many applications of plant pathology had been enabled by the evolution of artificial intelligence (AI). For instance, many researchers had used pre-trained convolutional neural networks (CNNs) such as the VGG-16, Inception, and Google Net to mention a few, for the classifications of plant diseases. The trend of using AI for plant disease classification has grown to such an extent that some researchers were able to use artificial intelligence to also detect their severities. The purpose of this study is to introduce a novel approach that is reliable in predicting severities of the maize common rust disease by CNN deep learning models. This was achieved by applying threshold-segmentation on images of diseased maize leaves (Common Rust disease) to extract the percentage of the diseased leaf area which was then used to derive fuzzy decision rules for the assignment of Common Rust images to their severity classes. The four severity classes were then used to train a VGG-16 network in order to automatically classify the test images of the Common Rust disease according to their classes of severity. Trained with images developed by using this proposed approach, the VGG-16 network achieved a validation accuracy of 95.63% and a testing accuracy of 89% when tested on images of the Common Rust disease among four classes of disease severity named Early stage, Middle stage, Late Stage and Healthy stage.
Highlights
Researchers who have so far used deep learning models to classify plant leaf disease severities have used training datasets that were categorized into classes that relied on human decisions by making observations
The botanist they used, made the following discretions to make decisions on severity classes: the healthy-stage leaves are free of spots; the early-stage leaves have small circular spots with diameters less than 5 mm; the middle-stage leaves have more than three spots with at least one frog-eye spot enlarging to irregular or lobed shape; the end-stage leaves are so heavily infected that they will drop from the tree. With these decisions that were made by their botanist, they developed the training datasets that were used to train the Visual Geometry Group (VGG)-16 network with a testing accuracy of 90.4%. The approach they used was not generic, in the sense that the methods used to assign diseased images to their severity classes were exclusive to apple leaf diseases, and the severity class assignments were made based on the observation of a human eye, in this study we introduce a novel approach that uses the decisions of computerized fuzzy decision rules to assign the maize Common Rust images to their severity classes of data sets that were used to train the VGG-16 network
This change in colour results in the loss of the green pigment of the leaves. This proposed approach was achieved by use of the Otsu threshold-segmentation method that was used to extract the percentages of the diseased leaf area (Background pixels) which were used to derive fuzzy decision rules for the assignment of maize Common Rust images to their severity classes
Summary
Researchers who have so far used deep learning models to classify plant leaf disease severities have used training datasets that were categorized into classes that relied on human decisions by making observations. The botanist they used, made the following discretions to make decisions on severity classes: the healthy-stage leaves are free of spots; the early-stage leaves have small circular spots with diameters less than 5 mm; the middle-stage leaves have more than three spots with at least one frog-eye spot enlarging to irregular or lobed shape; the end-stage leaves are so heavily infected that they will drop from the tree With these decisions that were made by their botanist, they developed the training datasets that were used to train the VGG-16 network with a testing accuracy of 90.4%. Once the severity classes were developed, we trained the fine-tuned VGG-16 network to classify the tested maize common rust images among four disease severity classes: Early Stage, Middle Stage, Late Stage, and Healthy Stage Using this approach, the RGB images of the maize Common Rust disease were first converted to grayscale.
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
