Industrial IoT Technologies and Applications. International Conference, Industrial IoT 2016, GuangZhou, China, March 25-26, 2016, Revised Selected Papers

Research Article

Junction Based Table Detection in Mobile Captured Golf Scorecard Images

Download137 downloads
  • @INPROCEEDINGS{10.1007/978-3-319-44350-8_18,
        author={Junying Yuan and Haishan Chen and Huiru Cao and Zhonghua Guo},
        title={Junction Based Table Detection in Mobile Captured Golf Scorecard Images},
        proceedings={Industrial IoT Technologies and Applications. International Conference, Industrial IoT 2016, GuangZhou, China, March 25-26, 2016, Revised Selected Papers},
        proceedings_a={INDUSTRIALIOT},
        year={2016},
        month={9},
        keywords={Mobile captured images Junction detection Table detection Pair-wise relationship Junction filtering Junction recovery},
        doi={10.1007/978-3-319-44350-8_18}
    }
    
  • Junying Yuan
    Haishan Chen
    Huiru Cao
    Zhonghua Guo
    Year: 2016
    Junction Based Table Detection in Mobile Captured Golf Scorecard Images
    INDUSTRIALIOT
    Springer
    DOI: 10.1007/978-3-319-44350-8_18
Junying Yuan1,*, Haishan Chen1, Huiru Cao1, Zhonghua Guo1
  • 1: Nanfang College of Sun Yat-Sen University
*Contact email: cihisa@outlook.com

Abstract

Table detection in mobile captured images faces many challenges owning to the well-known low image quality. Recently, a few researches pioneer in detecting the tables in rich-text images, but few works for scorecard images which usually lack of texts but are rich in graphics, such as golf scorecard images. In this paper, a junction-relation based table detection method for mobile captured scorecard images is proposed. Firstly, the most distinguished junctions are determined via a simplified pattern matching method, then the fault detections are removed through filtering operations, finally the missed junctions are recovered utilizing the pair-wise relationships among neighboring junctions. The experimental results show that 98.47 % of the junctions from 90 test images are correctly detected, and thus proves the superiority of the proposed method.