第13卷第3期 智能系统学报 Vol.13 No.3 2018年6月 CAAI Transactions on Intelligent Systems Jun.2018 D0:10.11992/tis.201612040 网络出版地址：http:/kns.cnki.net/kcms/detail/23.1538.TP.20170702.0425.018.html 高斯核函数卷积神经网络跟踪算法 汪鸿翔，柳培忠'，骆炎民2，杜永兆，陈智 (1.华侨大学工学院.福建泉州362021：2.华侨大学计算机科学与技术学院，福建厦门361021) 摘要：针对深度学习跟踪算法训练样本缺少、训练费时、算法复杂度高等问题，引入高斯核函数进行加速，提出一种 无需训练的简化卷积神经网络跟踪算法。首先，对初始帧目标进行归一化处理并聚类提取一系列初始滤波器组，跟 踪过程中结合目标背景信息与前景候选目标进行卷积：然后，提取目标简单抽象特征：最后，将简单层的卷积结果进 行叠加得到目标的深层次特征表达。通过高斯核函数加速来提高算法中全部卷积运算的速度，利用目标的局部结构 特征信息，对网络各阶段滤波器进行更新，结合粒子滤波跟踪框架实现跟踪。在CVP℉2013跟踪数据集上的实验表 明，本文方法脱离了繁琐深度学习运行环境，能克服低分辨率下目标局部遮挡与形变等问题，提高复杂背景下的跟踪 效率。 关键词：视觉跟踪；深度学习；卷积神经网络；高斯核函数；前景目标；背景信息；模板匹配：粒子滤波 中图分类号：TP391文献标志码：A文章编号：1673-4785(2018)03-0388-07 中文引用格式：汪鸿翔，柳培忠，骆炎民，等.高斯核函数卷积神经网络跟踪算法J.智能系统学报，2018,13(3)：388-394 英文引用格式：VANG Hongxiang,LIU Peizhong,LUO Yanmin,etal.Convolutional neutral network tracking algorithm acceler-. ated by Gaussian kernel function[J.CAAI transactions on intelligent systems,2018,13(3):388-394. Convolutional neutral network tracking algorithm accelerated by Gaussi- an kernel function WANG Hongxiang',LIU Peizhong',LUO Yanmin',DU Yongzhao',CHEN Zhi' (1.College of Engineering,Huaqiao University,Quanzhou 362021,China;2.College of Computer Science and Technology,Huaqiao University,Xiamen 361021,China) Abstract:In view of such defects existing in the depth learning tracking algorithm as lack of training samples,large time consumption,and high complexity,this paper proposed a simplified convolutional neural network tracking al- gorithm in which training is unnecessary.Moreover,the Gaussian kernel function can be applied to this algorithm to sig- nificantly lower the computing time.Firstly,the initial frame target was normalized and clustered to extract a series of initial filter banks;in the tracking process,the background information of the target and the candidate target for the fore- ground were convoluted;then the simple and abstract features of the target were extracted;finally,all the convolutions of a simple layer were superposed to form a deep-level feature representation.The Gaussian kernel function was used to speed-up the convolution operations:also,the local structural feature information of the target was used to update the fil- ters in every stage of the network;in addition,the tracking was realized by combining the particle filter tracking frame- work.The experimental results on the CVPR2013 tracking datasets show that the method used in this paper can help avoid the typically cumbersome operational environment of deep learning,overcome local object occlusion and deform- ation at low resolution,and improve tracking efficiency under a complex background. Keywords:visual tracking;deep learning,convolutional neural network(CNN);gauss kernel function;foreground ob- ject,background information;template matching;particle filter 收稿日期：2016-12-31.网络出版日期：2017-07-02. 基金项目：国家自然科学基金项目(61203242,61605048)：福建省 视觉跟踪是计算机视觉领域的研究热点，在虚 自然科学基金项目(2016J01300,2015J01256):华侨大 学研究生科研创新能力培育计划资助项目(1511422004). 拟现实、人机交互、智能监控、增强现实、机器感知 通信作者：柳培忠.E-mail:pzliu@hqu.edu.cn. 等场景中有着重要的研究与应用价值。视觉跟踪通DOI: 10.11992/tis.201612040 网络出版地址: http://kns.cnki.net/kcms/detail/23.1538.TP.20170702.0425.018.html 高斯核函数卷积神经网络跟踪算法 汪鸿翔1 ，柳培忠1 ，骆炎民2 ，杜永兆1 ，陈智1 （1. 华侨大学 工学院，福建 泉州 362021; 2. 华侨大学 计算机科学与技术学院，福建 厦门 361021） 摘 要：针对深度学习跟踪算法训练样本缺少、训练费时、算法复杂度高等问题，引入高斯核函数进行加速，提出一种 无需训练的简化卷积神经网络跟踪算法。首先，对初始帧目标进行归一化处理并聚类提取一系列初始滤波器组，跟 踪过程中结合目标背景信息与前景候选目标进行卷积；然后，提取目标简单抽象特征；最后，将简单层的卷积结果进 行叠加得到目标的深层次特征表达。通过高斯核函数加速来提高算法中全部卷积运算的速度，利用目标的局部结构 特征信息，对网络各阶段滤波器进行更新，结合粒子滤波跟踪框架实现跟踪。在 CVPR2013 跟踪数据集上的实验表 明，本文方法脱离了繁琐深度学习运行环境，能克服低分辨率下目标局部遮挡与形变等问题，提高复杂背景下的跟踪 效率。 关键词：视觉跟踪；深度学习；卷积神经网络；高斯核函数；前景目标；背景信息；模板匹配；粒子滤波 中图分类号：TP391 文献标志码：A 文章编号：1673−4785(2018)03−0388−07 中文引用格式：汪鸿翔, 柳培忠, 骆炎民, 等. 高斯核函数卷积神经网络跟踪算法[J]. 智能系统学报, 2018, 13(3): 388–394. 英文引用格式：WANG Hongxiang, LIU Peizhong, LUO Yanmin, et al. Convolutional neutral network tracking algorithm acceler￾ated by Gaussian kernel function[J]. CAAI transactions on intelligent systems, 2018, 13(3): 388–394. Convolutional neutral network tracking algorithm accelerated by Gaussi￾an kernel function WANG Hongxiang1 ，LIU Peizhong1 ，LUO Yanmin2 ，DU Yongzhao1 ，CHEN Zhi1 (1. College of Engineering, Huaqiao University, Quanzhou 362021, China; 2. College of Computer Science and Technology, Huaqiao University, Xiamen 361021, China) Abstract: In view of such defects existing in the depth learning tracking algorithm as lack of training samples, large time consumption, and high complexity, this paper proposed a simplified convolutional neural network tracking al￾gorithm in which training is unnecessary. Moreover, the Gaussian kernel function can be applied to this algorithm to sig￾nificantly lower the computing time. Firstly, the initial frame target was normalized and clustered to extract a series of initial filter banks; in the tracking process, the background information of the target and the candidate target for the fore￾ground were convoluted; then the simple and abstract features of the target were extracted; finally, all the convolutions of a simple layer were superposed to form a deep-level feature representation. The Gaussian kernel function was used to speed-up the convolution operations; also, the local structural feature information of the target was used to update the fil￾ters in every stage of the network; in addition, the tracking was realized by combining the particle filter tracking frame￾work. The experimental results on the CVPR2013 tracking datasets show that the method used in this paper can help avoid the typically cumbersome operational environment of deep learning, overcome local object occlusion and deform￾ation at low resolution, and improve tracking efficiency under a complex background. Keywords: visual tracking; deep learning; convolutional neural network (CNN); gauss kernel function; foreground ob￾ject; background information; template matching; particle filter 视觉跟踪是计算机视觉领域的研究热点，在虚 拟现实、人机交互、智能监控、增强现实、机器感知 等场景中有着重要的研究与应用价值。视觉跟踪通 收稿日期：2016−12−31. 网络出版日期：2017−07−02. 基金项目：国家自然科学基金项目 (61203242，61605048)；福建省 自然科学基金项目 (2016J01300，2015J01256)；华侨大 学研究生科研创新能力培育计划资助项目 (1511422004). 通信作者：柳培忠. E-mail：pzliu@hqu.edu.cn. 第 13 卷第 3 期 智 能 系 统 学 报 Vol.13 No.3 2018 年 6 月 CAAI Transactions on Intelligent Systems Jun. 2018