A Study on the Efficiency of Recognition of Specific Workers based on Deep Learning through Wearing a Work Vest

초록

This paper studies the method of detecting and tracking a moving target (person) among workers. The target’s noticeable features, such as its shape or color, are some of the requirements that guarantee the reliable performance of human detection and tracking methods. These features, however, are not adequate for the target tracking methods among workers because they all have a similar shape. Moreover, their clothes vary from day to day. Accordingly, we overcame these issues by devising a method in which the target wears a work vest, thus creating a distinctive feature. The contribution of this paper is as follows: 1) A work vest can be used as a noticeable feature to detect and track a target among workers. 2) Even if the target changes, the proposed method can be performed without additional learning. 3) Finally, by implementing the prototype system, the possibility of the proposed methods was experimentally verified. © ICROS 2022.

키워드