|
|||||
|
|
| 基于Adaboost首帧检测的时空上下文人脸跟踪算法 | |
| 引用本文： | 张尧,才华,李心达,米晓红,孙俊喜.基于Adaboost首帧检测的时空上下文人脸跟踪算法[J].吉林大学学报(理学版),2020,58(2):314-320. |
| 作者姓名： | 张尧 才华 李心达 米晓红 孙俊喜 |
| 作者单位： | 1. 长春理工大学 电子信息工程学院, 长春 130022； 2. 长春中国光学科学技术馆, 长春 130117； 3. 河南科技大学 管理学院, 河南 洛阳 471023； 4. 东北师范大学 信息科学与技术学院, 长春 130117 |
| 基金项目： | 国家自然科学基金;吉林省科技发展计划 |
| 摘 要： | 针对时空上下文算法首帧需进行手动框选及选择偏差带来后续跟踪干扰的不足, 提出利用Adaboost算法进行首帧检测, 引入Kalman预测机制辅助时空上下文算法进行跟踪.当出现遮挡、 抖动等问题时, 保证跟踪稳定地进行, 提高算法鲁棒性. 在Shelter1等3组公共数据集上进行对比实验的结果表明, 该算法能实现首帧自动检测功能, 后续跟踪算法的鲁棒性及跟踪效果也得到明显提升. |
| 关 键 词： | 时空上下文 Adaboost算法 Kalman滤波 视觉跟踪 |
| 收稿时间： | 2019-04-25 |
Spatio Temporal Context Face Tracking AlgorithmBased on Adaboost First Frame Detection |
|
| ZHANG Yao,CAI Hua,LI Xinda,MI Xiaohong,SUN Junxi.Spatio Temporal Context Face Tracking AlgorithmBased on Adaboost First Frame Detection[J].Journal of Jilin University: Sci Ed,2020,58(2):314-320. | |
| Authors: | ZHANG Yao CAI Hua LI Xinda MI Xiaohong SUN Junxi |
| Institution: | 1. School of Electronic Information Engineering, Changchun University of Science and Technology, Changchun 130022, China； 2. Changchun China Optics Science and Technology Museum, Changchun 130117, China; 3. School of Management, Henan University of Science and Technology, Luoyang 471023, Henan Province, China; 4. School of Information Science and Technology, Northeast Normal University, Changchun 130117, China |
| Abstract: | Aiming at the problem that the first frame needed to be manually selected and the subsequent tracking interference was caused by selection deviations in the spatio temporal context tracking algorithm, we proposed to use Adaboost algorithm to detect the first frame and introduce Kalman prediction mechanism to assist spatio temporal context tracking algorithm. When occlusion, jitter and other problems occurred, it could ensure the stability of tracking and improve the robustness of the algorithm. Comparative experiments were carried out on three public data sets, such as Shelter1. The experimental results show that the proposed algorithm can realize the automatic detection function of the first frame, and the robustness and tracking effect of the subsequent tracking algorithm are also significantly improved. |
| Keywords: | spatio temporal context Adaboost algorithm Kalman filter visual tracking |
| 本文献已被 CNKI 万方数据 等数据库收录！ | |
| 点击此处可从《吉林大学学报(理学版)》浏览原始摘要信息 | |
| 点击此处可从《吉林大学学报(理学版)》下载免费的PDF全文 | |