Scholarlink

<p>The merging of artificial intelligence and industrial applications has proved its worth in many fields. Logistics is one area that could benefit from this innovation. In particular, automating the task of detecting the number of boxes on pallet significantly streamlines logistics and enhances the overall shipping experience for consumers. However, the high complexity of this task, combined with the lack of publicly available datasets depicting pallets in a warehouse, has limited advancements in the literature. To address this challenge, we propose a new lightweight framework for 2D/3D reconstruction, designed to count the number of missing boxes on a pallet. Our method efficiently infers the 3D positions of individual objects from a 2D image by reconstructing a sparse 3D point cloud. This object-centric approach prioritizes computational efficiency and practical utility, enabling accurate counting of missing boxes on the pallet, even in scenarios involving occluded objects. Moreover, to overcome the lack of open databases in the logistics domain, we release a new synthetic dataset designed to simulate boxes on a pallet within a warehouse setting. Extensive experiments demonstrate the ability of our proposed framework in accurately reconstructing the exact position of each box and consequently counting the number of missing boxes on a pallet with an accuracy of <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="10489_2025_6621_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="37" /> </InlineMediaObject> <EquationSource Format="TEX">\(\varvec{98\%}\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn mathvariant="bold">98</mn> <mo mathvariant="bold">%</mo> </mrow> </math></EquationSource> </InlineEquation>. The code is available at <a href="https://github.com/ikramedd/-3D-PRNet--2D-3D-Reconstruction">https://github.com/ikramedd/-3D-PRNet--2D-3D-Reconstruction</a>.</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

Towards automation of warehouse management: Counting boxes on pallets via 3D reconstruction from a single image

  • 期刊论文

  • 发表日期：

    2025年6月18日

查看全文
  • Ikram Eddahmani,
  • Thibault Napoléon,
  • Chi-Hieu Pham,
  • Isabelle Badoc,
  • Marwa EL-Bouz

摘要

The merging of artificial intelligence and industrial applications has proved its worth in many fields. Logistics is one area that could benefit from this innovation. In particular, automating the task of detecting the number of boxes on pallet significantly streamlines logistics and enhances the overall shipping experience for consumers. However, the high complexity of this task, combined with the lack of publicly available datasets depicting pallets in a warehouse, has limited advancements in the literature. To address this challenge, we propose a new lightweight framework for 2D/3D reconstruction, designed to count the number of missing boxes on a pallet. Our method efficiently infers the 3D positions of individual objects from a 2D image by reconstructing a sparse 3D point cloud. This object-centric approach prioritizes computational efficiency and practical utility, enabling accurate counting of missing boxes on the pallet, even in scenarios involving occluded objects. Moreover, to overcome the lack of open databases in the logistics domain, we release a new synthetic dataset designed to simulate boxes on a pallet within a warehouse setting. Extensive experiments demonstrate the ability of our proposed framework in accurately reconstructing the exact position of each box and consequently counting the number of missing boxes on a pallet with an accuracy of \(\varvec{98\%}\) 98 % . The code is available at https://github.com/ikramedd/-3D-PRNet--2D-3D-Reconstruction.