摘要
目标识别和定位是计算机视觉领域研究的主要问题,图像分割、目标跟踪、目标行为分析等都是以图像中的目标检测为基础的。随着深度学习技术的发展,目标检测算法取得了巨大突破。在广泛调研相关文献的基础上,对目标检测算法进行分析和对比,分别研究基于区域提取的两阶段目标检测架构和直接位置回归的一阶段目标检测架构的本质特点和发展过程,并提出未来的发展方向。
Target recognition and localization are the main research issues in the field of computer vision.Image segmentation,target tracking,target behavior analysis are all based on target detection in images.With the development of deep learning technology,target detection algorithm has made a great breakthrough.On the basis of extensive research on relevant literature,this paper analyzes and compares the target detection algorithm,respectively studies the essential characteristics and development process of two-stage target detection architecture based on region extraction and one-stage target detection architecture based on direct location regression,and puts forward the direction for future development.
作者
陈辉东
丁小燕
刘艳霞
CHEN Huidong;DING Xiaoyan;LIU Yanxia(College of Urban Rail Transit and Logistics,Beijing Union University,Beijing 100101,China;Institute of Geographical Sciences,Hebei Academy of Sciences,Shijiazhuang 050011,China)
出处
《北京联合大学学报》
CAS
2021年第3期39-46,共8页
Journal of Beijing Union University
基金
北京市自然科学基金项目(3214062)
北京联合大学人才强校优选-拔尖计划项目(BPHR2020BZ02)
北京市教委科技计划项目(KM202111417004,KM202011417004,KM201911417007)
北京联合大学科研项目(ZK30202002)。
关键词
目标检测
深度学习
卷积神经网络
Target detection
Deep learning
Convolutional neural networks
