AI pipeline sharpens detection of yellow rust on wheat leaves

Yellow rust, a fungal disease of wheat, can seriously damage crop production and threaten global food security. A new study in the International Journal of Data Science and Analytics reports a computer vision method that classifies different levels of yellow rust infection on wheat leaves, with the aim of supporting timely disease control in the field.

The research team, led by M. Chilakarao with Santi Kumari Behera and Prabira Kumar Sethy, combined a novel image enhancement stage with deep learning feature extraction. The enhancement uses two-stage contrast improvement guided by Harris Hawks optimization, a nature-inspired search algorithm, together with a contrast preservation metric to keep the image faithful while boosting visual detail.

After enhancement, the system extracts features from wheat leaf images using three convolutional neural networks: ResNet50, EfficientNet-B0 and VGG16. The resulting feature sets are merged through a weighted average, and Principal Component Analysis reduces the dimensionality to form an informative feature set for classification.

Classification is then performed by a Support Vector Machine whose hyperparameters are optimized using Bayesian optimization. This staged pipeline moves from raw image to enhanced image to fused deep features to a final decision about the extent of yellow rust on a leaf.

The method was tested on the YELLOW-RUST-19 dataset, which contains 15,000 images showing different levels of yellow rust. According to the authors, classification accuracy rose from 78 percent before enhancement to 87 percent after enhancement, with an area under the curve of 0.9761.

The authors state that the results show improved image quality of visual features using the suggested system.

Because the evaluation was carried out on a single benchmark dataset of leaf images, the findings describe performance under those test conditions rather than confirmed performance in commercial farm deployments. The authors note that the same approach can be utilized for building intelligent disease prediction systems for precision agriculture in the field.

Timely classification of the extent of disease on leaves is necessary for efficient disease control, the paper notes, and automated systems of this kind could help farmers identify infection levels earlier. The work appears as a regular paper in Volume 22 of the journal, published on 23 October 2026.

Subject of Research: Machine learning classification of yellow rust disease levels on wheat leaves using image enhancement and deep feature fusion

Article Title: Enhanced yellow rust classification on wheat leaves using Harris Hawks optimization-guided contrast enhancement and deep feature fusion for precision agriculture

Article References: Chilakarao, M., Behera, S. K., & Sethy, P. K. (2026). Enhanced yellow rust classification on wheat leaves using Harris Hawks optimization-guided contrast enhancement and deep feature fusion for precision agriculture. International Journal of Data Science and Analytics, 22(1), Article 352. https://doi.org/10.1007/s41060-026-01343-y

Image Credits: AI Generated

DOI: 10.1007/s41060-026-01343-y

Keywords: yellow rust, wheat, plant disease detection, deep learning, image enhancement, Harris Hawks optimization, ResNet50, EfficientNet-B0, VGG16, support vector machine, precision agriculture, computer vision