Enhanced wheat crop leaf disease classification using multi-level contrast enhancement and modified vision...

Area of Science:

• Agricultural Science
• Computer Vision
• Plant Pathology

Background:

• Wheat is crucial for global food security but vulnerable to diseases impacting yield and quality.
• Early and accurate disease detection is essential for effective management and yield optimization.

Purpose of the Study:

• To develop and evaluate an innovative approach for rapid and precise diagnosis of wheat crop leaf diseases.
• To combine multi-level contrast enhancement with a novel transformer-based architecture for improved disease detection.

Main Methods:

• Utilized a multi-level contrast enhancement framework to improve wheat crop image quality.
• Employed Vision Transformers (ViTs) for efficient, multi-scale feature extraction and dimensionality reduction.
• Evaluated three ViT variants (modified seven-block, ViT-16-Tiny, modified seven-block with skip connections) on public wheat datasets.

Main Results:

• Achieved high classification accuracies: 98.90% (modified seven-block ViT), 97.50% (ViT-16-Tiny), and 97.90% (modified seven-block with skip connections).
• Demonstrated superior performance compared to state-of-the-art methods in accuracy, precision, sensitivity, and False Negative Rate.
• Showcased reduced total learnable parameters compared to existing techniques.

Conclusions:

• The proposed contrast enhancement and ViT-based approach offers a highly accurate and efficient solution for wheat disease diagnosis.
• This AI-driven method has the potential to significantly enhance agricultural productivity and food security.
• The study highlights the effectiveness of advanced deep learning techniques in addressing critical challenges in crop disease management.