Scholarlink

<p>Aiming to achieve accurate and rapid pallet identification, an efficient pallet localization method is proposed in this study, which can autonomously identify pallets and determine their insertion locations without manual annotation, by utilizing on-site point cloud data. yolov5s model was employed to obtain the region of interest (ROI), through innovative elevation filter and other processing, the point cloud was extracted and subsequently registered with the template. In the process of identification and localization, our research takes into account the geometric characteristics of the object, which avoids the system mistakenly identifying the images of pallets printed on packaging as actual pallets, and improves the safety. Field experiments demonstrate that the system can rapidly achieve full-angle localization with an accuracy of 94 %, while the positioning time is less than 1.5 seconds. These results provide a feasible strategy for intelligent logistics systems.</p>

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

Study on 3D pallet recognition and localization based on a depth camera

  • 期刊论文

  • 发表日期：

    2025年12月10日

查看全文
  • Haixia Hu,
  • Cenyang Xiang,
  • Chen Kong

摘要

Aiming to achieve accurate and rapid pallet identification, an efficient pallet localization method is proposed in this study, which can autonomously identify pallets and determine their insertion locations without manual annotation, by utilizing on-site point cloud data. yolov5s model was employed to obtain the region of interest (ROI), through innovative elevation filter and other processing, the point cloud was extracted and subsequently registered with the template. In the process of identification and localization, our research takes into account the geometric characteristics of the object, which avoids the system mistakenly identifying the images of pallets printed on packaging as actual pallets, and improves the safety. Field experiments demonstrate that the system can rapidly achieve full-angle localization with an accuracy of 94 %, while the positioning time is less than 1.5 seconds. These results provide a feasible strategy for intelligent logistics systems.