Tomato Leafminer: "Early Detection And identification Tuta absoluta damage on tomato Using Hybrid Fuzzy PSPNet-SVM
الباحث الأول:
Adil Abdulhur Abushana
الباحثين الآخرين:
Hayfaa T. Hussein
المجلة:
International Journal of Informatics and Communication Technology (IJ-ICT)
تاريخ النشر:
21 إبريل، 2026
مختصر البحث:
The study introduces The PDRFSVM framework which combines PSPNet-based segmentation with multi-branch feature extraction from DenseNet121 and ResNet50V2 and a Vision Transformer (ViT) to achieve fuzzy classification through Fuzzy Support Vector Mach…
The study introduces The PDRFSVM framework which combines PSPNet-based segmentation with multi-branch feature extraction from DenseNet121 and ResNet50V2 and a Vision Transformer (ViT) to achieve fuzzy classification through Fuzzy Support Vector Machine (Fuzzy SVM). The ViT component uses self-attention to model long-range spatial dependencies which allows it to extract detailed lesion characteristics that include subtle discoloration patterns and irregular boundaries. The proposed framework combines heterogeneous deep features through structured fusion which enables it to effectively represent complex disease patterns across different environmental conditions. The experimental evaluation across various deep learning models shows that ResNet50V2 achieves strong baseline performance through optimization using both SGD and Adam while the PDRFSVM framework reaches higher accuracy of 98.2% which exceeds TLFDNN and standalone ResNet50V2 performance. Fuzzy SVM integration into the system provides a classification method that utilizes continuous membership values to create a grading system for disease intensity instead of using binary classification. The system helps precision agriculture by enabling farmers to detect problems early and create treatment strategies. The proposed approach presents a strong solution which detects Tuta absoluta and tomato leaf diseases at an early stage while assessing their severity through an interpretable system which agricultural systems can use in real-world applications.
