Scholarlink

The core functionality of any agricultural harvesting robot is its automated fruit detection system. Nevertheless, fruit detection is complicated by arduous environmental conditions, including illumination variance, occlusion from foliage, and the clustering of production. In fact, harvesting date fruit involves many risks, such as worker falls, because the palm trees are quite tall. Main challenge in automating this process is accurately identifying the date fruit bunch stalk, as the fruit cluster is attached to this stalk on the palm tree. In this study, a method for identifying and segmenting bunch stalks from photos is proposed using YOLOv8 model and SAM algorithm, reaching a 91.2% accuracy rate and a mean Average Precision (mAP50) of 95.6%. This technique can predict the bonding boxes coordinate of bunch stalk in image, and the SAM algorithm gives the mask of bunch stalk, that can be used to analyze color, size and texture of bunch stalk. We conclude that the YOLOv8 model can be effectively utilized to develop advanced computer vision systems that assist engineers in designing and deploying robotic or drone-based solutions for date palm fruit harvesting.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Detection and Segmentation of Date Fruit Bunch Stalk Using YOLOv8 and SAM Algorithms

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Youssef Bouh,
  • Lhoussaine Ait Ben Mouh,
  • Othmane Reddate,
  • Mohamed Ouhda

摘要

The core functionality of any agricultural harvesting robot is its automated fruit detection system. Nevertheless, fruit detection is complicated by arduous environmental conditions, including illumination variance, occlusion from foliage, and the clustering of production. In fact, harvesting date fruit involves many risks, such as worker falls, because the palm trees are quite tall. Main challenge in automating this process is accurately identifying the date fruit bunch stalk, as the fruit cluster is attached to this stalk on the palm tree. In this study, a method for identifying and segmenting bunch stalks from photos is proposed using YOLOv8 model and SAM algorithm, reaching a 91.2% accuracy rate and a mean Average Precision (mAP50) of 95.6%. This technique can predict the bonding boxes coordinate of bunch stalk in image, and the SAM algorithm gives the mask of bunch stalk, that can be used to analyze color, size and texture of bunch stalk. We conclude that the YOLOv8 model can be effectively utilized to develop advanced computer vision systems that assist engineers in designing and deploying robotic or drone-based solutions for date palm fruit harvesting.