Scholarlink

One of the most crucial objectives in modern robotics is the development of novel and accurate technologies for human detection and following in diverse environments. This work focuses on developing a general purpose template matching algorithm which can detect a walking stick held and maneuvered by a human being. Given that stick orientations and positions undergo various transformations during tracking, the stick detection problem can be formulated as a general-purpose template matching task. We utilize a modified version of the popular FAsT-Match algorithm, named Color FAsT-Match or CFAsT-Match for approximate template matching under 2D affine transformations. Further, this paper proposes a new clustering, named DENsity-based CLUstEring (DENCLUE), based affine template matching approach which overcomes typical disadvantages faced by the original DBSCAN (density-based spatial clustering of applications with noise) based CFAsT-Match. Performance evaluations demonstrate the superiority of the proposed affine color template matching with kernel density estimation based clustering (ACTM-KDEC) approach for walking stick detection during human tracking in human-robot coexisting environments.

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

Kernel Density Estimation Based Clustering for Visual Walking Stick Detection in Human-Robot Coexisting Environments

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Reeju Bhattacharya,
  • Ronojit Pal,
  • Pritam Paral,
  • Anindita Sengupta

摘要

One of the most crucial objectives in modern robotics is the development of novel and accurate technologies for human detection and following in diverse environments. This work focuses on developing a general purpose template matching algorithm which can detect a walking stick held and maneuvered by a human being. Given that stick orientations and positions undergo various transformations during tracking, the stick detection problem can be formulated as a general-purpose template matching task. We utilize a modified version of the popular FAsT-Match algorithm, named Color FAsT-Match or CFAsT-Match for approximate template matching under 2D affine transformations. Further, this paper proposes a new clustering, named DENsity-based CLUstEring (DENCLUE), based affine template matching approach which overcomes typical disadvantages faced by the original DBSCAN (density-based spatial clustering of applications with noise) based CFAsT-Match. Performance evaluations demonstrate the superiority of the proposed affine color template matching with kernel density estimation based clustering (ACTM-KDEC) approach for walking stick detection during human tracking in human-robot coexisting environments.