Early Detection of Apple Scab in Orchards and Mildew in Vineyards Using A Novel Deep Learning Model and Multispectral Imagery


Asiroglu B., Turkoglu A. K., Bir F. S., Senan S.

JOURNAL OF AGRICULTURAL SCIENCES-TARIM BILIMLERI DERGISI, cilt.32, sa.4, ss.1030-1050, 2026 (SCI-Expanded, Scopus, TRDizin)

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 32 Sayı: 4
  • Basım Tarihi: 2026
  • Doi Numarası: 10.15832/ankutbd.1898342
  • Dergi Adı: JOURNAL OF AGRICULTURAL SCIENCES-TARIM BILIMLERI DERGISI
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, BIOSIS, CAB Abstracts, Food Science & Technology Abstracts, Directory of Open Access Journals, TR DİZİN (ULAKBİM), Academic Search Ultimate (EBSCO)
  • Sayfa Sayıları: ss.1030-1050
  • Açık Arşiv Koleksiyonu: AVESİS Açık Erişim Koleksiyonu
  • İstanbul Üniversitesi-Cerrahpaşa Adresli: Evet

Özet

Agricultural diseases pose a significant threat to crop productivity by causing substantial damage to plant tissues and reducing yield. Therefore, accurate and early detection of plant diseases is critical for sustainable agricultural practices. In this study, a novel hybrid deep learning model is proposed for the detection and segmentation of plant diseases, specifically apple scab in orchards and mildew in vineyards. The proposed architecture integrates a U-Net framework with ConvLSTM to jointly model spatial, spectral, and temporal information from multispectral drone imagery. Unlike conventional U-Net + LSTM approaches, the model is adapted to process five-channel multispectral data and explicitly capture temporal dependencies across sequential observations, enabling improved detection of early-stage disease patterns. Multispectral time-series data were collected using a DJI Phantom 4 Multispectral drone over agricultural fields in Tekirdag, Turkey. The proposed model achieved an average Intersection over Union (IoU) of 82.05% for apple orchards and 79.67% for vineyards, demonstrating robust segmentation performance under real-world conditions. These results highlight the effectiveness of integrating multispectral and temporal information within a unified deep learning framework for accurate and timely plant disease detection in precision agriculture.