摘要
针对传统核相关滤波目标跟踪算法跟踪精确度不足问题,论文利用卷积神经网络来提取图像深度特征,克服传统特征的不鲁棒性。其次,结合改进核相关滤波目标跟踪算法,当目标发生遮挡时,能够准确对目标进行跟踪。论文选取代表性视频对算法进行了测试,并与传统特征进行了对比。结果显示该算法精确度提高17.6,算法跟踪速率达到20.59fps。
In order to solve the problem that the tracking accuracy of the traditional target tracking algorithm is insufficient, this paper uses the convolution neural network to extract the image depth features and overcome the robustness of the traditional fea. tures. Secondly,combined with the improved target tracking algorithm of nuclear correlation filtering,when the target is blocked, the target can be tracked accurately. In this paper,representative video is selected to test the algorithm and compared with tradition. al features. The results show that the accuracy of the algorithm is increased by 17.6 and the tracking rate reaches 20.59fps.
作者
梁华刚
高冬梅
庞丽琴
LIANG Huagang;GAO Dongmei;PANG Liqin(School of Electronic and Control Engineering,Chang'an University,Xi'an 710064)
出处
《计算机与数字工程》
2019年第5期1115-1119,共5页
Computer & Digital Engineering
基金
国家自然科学基金项目"基于高阶上下文建构学习的多源异质视频异常检测"(编号:61603057)
陕西省重点产业链-工业领域项目(编号:2017ZDL-G-2-3)资助
关键词
卷积神经网络
特征提取
相关滤波
目标跟踪
convolution neural network
feature extraction
correlation filtering
target tracking
