计算机工程与应用 ›› 2019, Vol. 55 ›› Issue (10): 16-29.DOI: 10.3778/j.issn.1002-8331.1810-0170

• 热点与综述 • 上一篇    下一篇

基于深度学习的视频跟踪研究进展综述

戴凤智,魏宝昌,欧阳育星,金  霞   

  1. 天津科技大学 电子信息与自动化学院,天津 300222
  • 出版日期:2019-05-15 发布日期:2019-05-13

Survey of Research Progress of Video Tracking Based on Deep Learning

DAI Fengzhi, WEI Baochang, OUYANG Yuxing, JIN Xia   

  1. School of Electronic Information and Automation, Tianjin University of Science and Technology, Tianjin 300222, China
  • Online:2019-05-15 Published:2019-05-13

摘要: 近年来深度学习迅猛发展,颠覆了语音识别、图像分类、文本理解等领域的算法设计思路。深度学习因其具备强大的特征提取能力,在图像识别领域的成绩尤为突出。然而深度学习与视频监控领域的结合并不多,由于深度模型具有多层网络结构,算法复杂度大,训练和更新模型时比较耗时,很难满足实时性要求。回顾了深度学习的发展史,介绍了最近10年来国内外深度学习主要模型,论述了基于深度学习的目标跟踪算法,指出了各算法的优缺点,最后对当前该领域存在的问题和发展前景进行了总结和展望。

关键词: 深度学习, 视频跟踪, 卷积神经网络, 递归神经网络, 自编码器

Abstract: In recent years, deep learning has developed rapidly. It has overturned the design of algorithms for speech recognition, image classification and text understanding. Because of its strong feature extraction ability, deep learning has achieved outstanding results in the field of image recognition. However, the combination of deep learning and video surveillance is not much. Because the depth model has a multi-layer network structure and the complexity of the algorithm is large, it takes time to train and update the model, and it is difficult to meet the real-time requirements. This paper reviews the development history of deep learning, and introduces the main model of deep learning in the last 10 years. Then, this paper discusses the target tracking algorithm based on deep learning and points out the advantages and disadvantages of each algorithm. Finally, the current problems and development prospects in this field are summarized and prospected.

Key words: deep learning, video tracking, convolutional neural network, recurrent neural network, autoencoder