inis 22(31): 1

Research Article

Application of Computer Vision in T-Shirt Dimensions Measurement

Download15 downloads
  • @ARTICLE{10.4108/eetinis.v9i31.707,
        author={Ngoc-Bich Le and Thi-Thu-Hien Pham and Quoc-Hung Phan and Narayan C. Debnath and Ngoc-Huan Le},
        title={Application of Computer Vision in T-Shirt Dimensions Measurement},
        journal={EAI Endorsed Transactions on Industrial Networks and Intelligent Systems},
        volume={9},
        number={31},
        publisher={EAI},
        journal_a={INIS},
        year={2022},
        month={4},
        keywords={T-Shirt, Auto-dimensioning, computer vision, metrology automation, intelligent system},
        doi={10.4108/eetinis.v9i31.707}
    }
    
  • Ngoc-Bich Le
    Thi-Thu-Hien Pham
    Quoc-Hung Phan
    Narayan C. Debnath
    Ngoc-Huan Le
    Year: 2022
    Application of Computer Vision in T-Shirt Dimensions Measurement
    INIS
    EAI
    DOI: 10.4108/eetinis.v9i31.707
Ngoc-Bich Le1, Thi-Thu-Hien Pham1, Quoc-Hung Phan2, Narayan C. Debnath3, Ngoc-Huan Le3,*
  • 1: Ho Chi Minh City International University
  • 2: National United University
  • 3: Eastern International University
*Contact email: Huan.le@eiu.edu.vn

Abstract

This paper presents a solution to automatically measure the T-shirt dimensions in the garment industry. To address this goal, the paper focuses on utilizing image processing to determine the T-shirt's dimensions. The processing algorithm was provided along with the proposed recognition regions novel approach that was expected to deliver faster processing speed and enhance accuracy. The feasibility was demonstrated by characterizing the accuracy and processing speed. Specifically, five distinctive dimensions were successfully identified and measured; with the replication of 30, the discrepancy varies from 0.095% (for chest) to 2.088% (for collar). The divergence is insignificant compared with the granted tolerances. Finally, the processing time and the mechanical structure of the system deliver productivity of 22 products/minute which is approximately 10 times more rapidly than manual measurement (25 seconds).