EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Deep Learning for Ripeness Grading of Oil Palm Fresh Fruit Bunches: A Comprehensive Review of Convolutional Neural Network Approaches

Wahyu Nurkholis Hadi Syahputra, Chatchawan Chaichana, Damorn Bundhurat, Patiwet Wuttisarnwattana, Bayu Taruna Widjaja Putra

Trends in Sciences · 30.03.2026 · 23 (8) · 13227

DOI: 10.48048/tis.2026.13227 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Trends in Sciences, online erschienen am 30.03.2026. Verfasst von Wahyu Nurkholis Hadi Syahputra und 4 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Palm oil is a strategic agricultural crop in Indonesia, Malaysia, and Thailand, contributing significantly to national economies and requiring continuous improvements in harvesting efficiency and mill operations. The growing demand for higher efficiency and consistent quality in palm oil mills has accelerated the adoption of advanced technologies, particularly artificial intelligence (AI), which is increasingly applied across agricultural sectors, including oil palm production. This review aims to examine the development of convolutional neural network (CNN)-based approaches for ripeness grading of oil palm fresh fruit bunches (FFB) using CNN techniques. It provides an overview of research trends and technical progress in this field, showing that Malaysia leads scientific publications related to palm oil ripeness detection, followed by Indonesia. Most existing studies employ 1-stage object detectors, especially YOLO-based architectures, due to their real-time capability and relatively high performance. However, these methods are often trained and evaluated using datasets limited to specific environments, plantation conditions, or fruit varieties, which constrains generalization and large-scale deployment. Key research gaps are identified, including limited dataset diversity, high computational requirements, insufficient integration with Internet of Things (IoT)–based plantation and mill management systems, and the lack of real-time estimation of quality indicators such as free fatty acid (FFA) content and kernel-related attributes. Future research directions highlight the need for multimodal sensing, multi-camera systems, and multi-task learning frameworks that integrate ripeness grading with oil extraction rate (OER) estimation to support more effective operational decision-making in palm oil production systems. HIGHLIGHTS Research on CNN in palm oil has increased since 2017 with the adoption of DL. Malaysia leads publications in palm oil ripeness research, followed by Indonesia and Thailand. One-stage detectors, particularly YOLO-based models, are most commonly used for real-time grading. Most CNN-based systems rely on external visual features for ripeness classification. Multimodal approaches for real-time biochemical assessment remain limited. GRAPHICAL ABSTRACT

Zitieren

APA

Syahputra, W. N. H., Chaichana, C., Bundhurat, D., Wuttisarnwattana, P., & Putra, B. T. W. (2026). Deep Learning for Ripeness Grading of Oil Palm Fresh Fruit Bunches: A Comprehensive Review of Convolutional Neural Network Approaches. Trends in Sciences, 23(8), 13227. https://doi.org/10.48048/tis.2026.13227

Vancouver

Syahputra WNH, Chaichana C, Bundhurat D, Wuttisarnwattana P, Putra BTW. Deep Learning for Ripeness Grading of Oil Palm Fresh Fruit Bunches: A Comprehensive Review of Convolutional Neural Network Approaches. Trends in Sciences. 2026;23(8):13227. doi:10.48048/tis.2026.13227

BibTeX
@article{Syahputra2026,
  author = {Syahputra, Wahyu Nurkholis Hadi and Chaichana, Chatchawan and Bundhurat, Damorn and Wuttisarnwattana, Patiwet and Putra, Bayu Taruna Widjaja},
  title = {Deep Learning for Ripeness Grading of Oil Palm Fresh Fruit Bunches: A Comprehensive Review of Convolutional Neural Network Approaches},
  journal = {Trends in Sciences},
  year = {2026},
  volume = {23},
  number = {8},
  pages = {13227},
  doi = {10.48048/tis.2026.13227}
}
Kostenlos starten →