Share
Export Citation
Siam Citrus Disease Detection Based on Aerial Images Using Adaptive Gamma Correction and Single-Scale Retinex
Asriani
2026 IEEE International Conference on Artificial Intelligence and Mechatronics Systems Aims 2026
Abstract
Siam citrus is one of the horticultural commodities with an important contribution to national production, but the production of this commodity faces serious challenges due to disease attacks, such as citrus vein phloem degeneration (CVPD) and citrus canker, which have been shown to significantly reduce productivity and crop quality. Therefore, more adaptive technology that can handle visual challenges in dynamic field environments is needed. The method used is YOLOv12 with an adaptive gamma correction approach to adjust lighting levels in the image contextually, as well as single scale retinex (SSR), which increases contrast and clarifies the color features of leaves and fruits under natural field lighting conditions. This study applies multiclass classification to Siam orange leaves and fruits, enabling the model to recognize multiple disease categories simultaneously. Baseline evaluation results showed relatively limited performance, with a precision of 67%, a recall of 70%, and an mAP50 of 62%. After applying the proposed method, the model achieved a precision of 80%, a recall of 77%, and an mAP50 of 74%, indicating a significant improvement in disease detection performance. These results demonstrate that the proposed method enhances the model’s robustness under varying lighting conditions and complex backgrounds.