陈立琳1, 3,,,
余旺盛2,
马素刚1, 3,
范九伦1
1.西安邮电大学计算机学院 西安 710121
2.空军工程大学信息与导航学院 西安 710077
3.西安邮电大学陕西省网络数据分析与智能处理重点实验室 ??西安 ??710121
基金项目:国家自然科学基金(61473309, 61703423)
详细信息
作者简介:侯志强:男,1973年生,教授,博士生导师,研究方向为图像处理、计算机视觉
陈立琳:女,1989年生,硕士生,研究方向为计算机视觉、目标跟踪和深度学习
余旺盛:男,1985年生,博士,研究方向为计算机视觉、图像处理,模式识别
马素刚:男,1982年生,博士生,研究方向为计算机视觉、机器学习
范九伦:男,1964年生,教授,博士生导师,研究方向为模式识别、图像处理
通讯作者:陈立琳 454525999@qq.com
中图分类号:TP391.4计量
文章访问数:2704
HTML全文浏览量:1106
PDF下载量:81
被引次数:0
出版历程
收稿日期:2018-11-06
修回日期:2019-05-29
网络出版日期:2019-06-12
刊出日期:2019-09-10
Robust Visual Tracking Algorithm Based on Siamese Network with Dual Templates
Zhiqiang HOU1, 3,Lilin CHEN1, 3,,,
Wangsheng YU2,
Sugang MA1, 3,
Jiulun FAN1
1. Institute of Computer, Xi’an University of Posts and Telecommunications, Xi’an 710121, China
2. Information and Navigation Institute, Air Force Engineering University, Xi’an 710077, China
3. Shaanxi Key Laboratory of Network Data Analysis and Intelligent Processing, Xi’an University of Posts and Telecommunications, Xi’an 710121, China
Funds:The National Natural Science Foundation of China (61473309, 61703423)
摘要
摘要:近年来,Siamese网络由于其良好的跟踪精度和较快的跟踪速度,在视觉跟踪领域引起极大关注,但大多数Siamese网络并未考虑模型更新,从而引起跟踪错误。针对这一不足,该文提出一种基于双模板Siamese网络的视觉跟踪算法。首先,保留响应图中响应值稳定的初始帧作为基准模板R,同时使用改进的APCEs模型更新策略确定动态模板T。然后,通过对候选目标区域与2个模板匹配度结果的综合分析,对结果响应图进行融合,以得到更加准确的跟踪结果。在OTB2013和OTB2015数据集上的实验结果表明,与当前5种主流跟踪算法相比,该文算法的跟踪精度和成功率具有明显优势,不仅在尺度变化、平面内旋转、平面外旋转、遮挡、光照变化情况下具有较好的跟踪效果,而且达到了46 帧/s的跟踪速度。
关键词:Siamese网络/
目标跟踪/
双模板/
模板更新
Abstract:In recent years, the Siamese networks has drawn great attention in visual tracking community due to its balanced accuracy and speed. However, most Siamese networks model are not updated, which causes tracking errors. In view of this deficiency, an algorithm based on the Siamese network with double templates is proposed. First, the base template R which is the initial frame target with stable response map score and the dynamic template T which is using the improved APCEs model update strategy to determine are kept. Then, the candidate targets region and the two template matching results are analyzed, meanwhile the result response maps are fused, which could ensure more accurate tracking results. The experimental results on the OTB2013 and OTB2015 datasets show that comparing with the 5 current mainstream tracking algorithms, the tracking accuracy and success rate of the proposed algorithm are superior. The proposed algorithm not only displays better tracking effects under the conditions of scale variation, in-plane rotation, out-of-plane rotation, occlusion, and illumination variation, but also achieves real-time tracking by a speed of 46 frames per second.
Key words:Siamese network/
Object tracking/
Dual templates/
Template update
PDF全文下载地址:
https://jeit.ac.cn/article/exportPdf?id=681ba770-19f8-460d-a819-5773750ff715