|
|
 |
| |
|
|
|
Deep Belief Network for Citrus Leaf Disease Detection Using Hyperspectral Images |
|
|
|
PP: 1207-1217 |
|
|
|
doi:10.18576/amis/190519
|
|
|
|
Author(s) |
|
|
|
E. Aswini,
C. Vijayakumaran,
|
|
|
|
Abstract |
|
|
| Citrus plants are essential for the agricultural industry but are susceptible to various diseases that can cause significant economic losses. This paper proposes a method to detect citrus leaf diseases using a deep belief network (DBN) and hyperspectral imaging. Hyperspectral imaging provides a rich spectral and spatial information source to identify different types of citrus leaf diseases. We pre-process the Hyperspectral images by normalizing the data and reducing the dimensionality using Principal Component Analysis (PCA). We then train a DBN with multiple Restricted Boltzmann Machines (RBMs) layers on the pre-processed data. The DBN can learn complex patterns in the data and extract features for classification with 1000 citrus leaf images, which include healthy leaves and leaves with three different types of diseases. According to the findings of our investigation, our approach obtains an accuracy of 94.5% on the test dataset, which is superior to the performance of a number of other machine learning methods. Our approach could automate citrus leaf disease detection in the agricultural industry, improving crop yields and reducing economic losses.
|
|
|
|
|
|
|
|
