作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2019, Vol. 45 ›› Issue (12): 249-256,262. doi: 10.19678/j.issn.1000-3428.0053838

• 多媒体技术及应用 • 上一篇    下一篇

一种基于背景加权的多特征融合目标跟踪算法

龚红1a, 杨发顺1b, 王代强2, 丁召1b   

  1. 1. 贵州大学 a. 人民武装学院;b. 大数据与信息工程学院, 贵阳 550025;
    2. 贵州民族大学 机械电子工程学院, 贵阳 550025
  • 收稿日期:2019-01-28 修回日期:2019-03-11 发布日期:2019-03-14
  • 作者简介:龚红(1977-),女,讲师、博士研究生,主研方向为目标跟踪、图像与视频处理;杨发顺,副教授、博士;王代强,教授、博士;丁召(通信作者),教授、博士、博士生导师。
  • 基金资助:
    国家自然科学基金(61464002,11564005);贵州省教育厅创新群体重大研究项目(黔科合字[2017]035)。

A Multi-Feature Fusion Target Tracking Algorithm Based on Background Weighting

GONG Hong1a, YANG Fashun1b, WANG Daiqiang2, DING Zhao1b   

  1. 1a. People's Armed College;1b. College of Big Data and Information Engineering, Guizhou University, Guiyang 550025, China;
    2. School of Mechatronics Engineering, Guizhou Minzu University, Guiyang 550025, China
  • Received:2019-01-28 Revised:2019-03-11 Published:2019-03-14

摘要: 针对均值漂移(MS)目标跟踪算法受背景环境变化干扰较大的问题,提出一种基于背景加权的多特征融合目标跟踪算法BWMMS。引入基于目标模型与目标周围背景模型差分的加权函数,细化各像素对准确描述目标的重要程度,从而提高目标模板的分辨能力。结合颜色与纹理特征进行目标跟踪,构建基于目标和目标背景区域的特征自适应融合机制,使BWMMS算法能够根据跟踪场景变化自适应调整颜色与纹理特征的权值。实验结果表明,与MS算法、HRBW算法相比,该算法对环境变化的适应性较好,能取得更鲁棒的跟踪结果,且跟踪成功率高达94.84%。

关键词: 背景加权, 特征融合, 目标跟踪, 颜色特征, 纹理特征

Abstract: As the Mean Shift(MS) target tracking algorithm is subject to the changes of background environment,this paper proposes a multi-feature fusion target tracking algorithm BWMMS based on background weighting.First,this paper introduces the weighted function based on the differences between the target model and the background model around the target,and refines the importance of each pixel to the target description,so as to improve the resolution of the target template.Then this paper combines the color and texture to track the target.According to the target and the target background area,this paper builds a feature adaptive fusion mechanism,thus enabling the BWMMS algorithm to adaptively adjust the weight of color and texture features according to the changes of tracking scenes.The experimental results show that compared with the MS algorithm and HRBW algorithm,the proposed method has better adaptability to environment changes.Besides,it can achieve more robust tracking results,and the tracking success rate reaches 94.84%.

Key words: background weighting, feature fusion, target tracking, color feature, texture feature

中图分类号: