Evaluating Histogram Matching for Robust Deep learning-Based Grapevine Disease Detection

2026-04-21

AI Summary

This study addresses the limited generalization of grape disease detection models under complex field lighting conditions. To mitigate this issue, the authors propose a two-stage histogram matching (HM) integration strategy: the first stage standardizes input images to reduce domain shift, while the second stage incorporates HM as a controllable data augmentation technique to enhance training diversity. Evaluated on a dataset of 1,469 RGB images—including both leaf close-ups and canopy scenes—using a ResNet-18 backbone, the proposed method significantly improves model robustness in real-world, variable lighting environments. Notably, performance gains are especially pronounced in canopy-level scenarios, demonstrating the effectiveness of the approach in challenging field conditions.

Technology Category

• Computer Vision: Other Foundations of Computer Vision
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Variability in illumination is a primary factor limiting deep learning robustness for field-based plant disease detection. This study evaluates Histogram Matching (HM), a technique that transforms the pixel intensity distribution of an image to match a reference profile, to mitigate this in grapevine classification, distinguishing among healthy leaves, downy mildew, and spider mite damage. We propose a dual-stage integration of HM: (i) as a preprocessing step for normalization, and (ii) as a data augmentation technique to introduce controlled training variability. Experiments using 1,469 RGB images (comprising homogeneous leaf-focused and heterogeneous canopy samples) to train ResNet-18 models demonstrate that this combination significantly enhances robustness on real-world canopy images. While leaf-focused samples showed marginal gains, the canopy subset improved markedly, indicating that balancing normalization with histogram-based diversification effectively bridges the domain gap caused by uncontrolled lighting.

Problem

• illumination variability
• deep learning robustness
• grapevine disease detection
• domain gap
• field-based plant disease

Innovation

• Histogram Matching
• Domain Gap
• Data Augmentation
• Illumination Normalization
• Robust Deep Learning