摘要
运动目标的检测与跟踪是智能交通、智能监控等领域的重要组成部分,尤其是行人、车辆的检测与跟踪,对于行人行车安全、流量监控等都有着重要的意义。由于复杂交通场景中多目标检测与跟踪的难度较高,人工设计的目标特征无法满足复杂的多目标检测与跟踪。因此,本文提出一种基于堆栈式去噪自编码器深度学习框架的多目标检测与跟踪算法,利用海量的图片数据集训练深度学习模型,提取目标深度特征,用于多目标检测与跟踪。实验结果表明,基于堆栈式去噪自编码器深度学习框架的多目标检测与跟踪算法,提高了多目标检测的准确性,实现了更加鲁棒的多目标跟踪效果。
Moving target detection and tracking is an important part of intelligent transportation,intelligent monitoring and other fields.Especially the detection and tracking of pedestrians and vehicles are of great significance for pedestrian traffic safety and traffic monitoring.Due to the difficulty of multi-target detection and tracking in complex traffic scenes,the target features designed manually cannot meet the requirements of complex multi-target detection and tracking.Therefore,this paper proposes a multi-target detection and tracking algorithm based on stack de-noising self-encoder depth learning framework.Using the massive picture data set to train the depth learning model to extract the depth feature of the target,which is used in the multi-target detection and tracking.The experimental results showthat the multi-target detection and tracking algorithm based on the stack de-noising self-encoder depth learning framework improves the accuracy of the multi-target detection and achieves a more robust multi-target tracking effect.
作者
左国才
苏秀芝
陈明丽
张珏
吴小平
ZUO Guocai;SU Xiuzhi;CHEN Mingli;ZHANG Jue;WU Xiaoping(School of Software and Information Engineering,Hunan Software Vocational Institute,Xiangtan Hunan 411100,China;Xinhua Chuyi Industrial School,Xinhua Hunan 417600,China;Rail Transit and Communication Technology College,Hunan Railway Vocational and Technical College,Zhuzhou Hunan 412001,China)
出处
《智能计算机与应用》
2020年第7期203-205,共3页
Intelligent Computer and Applications
基金
湖南省自然科学基金(2020JJ7007)
湘潭市科技局指导性计划项目(ZDX-CG2019013)
关键词
深度学习
堆栈式去噪自编码器
多目标检测与跟踪
Deep Learning
Stacked Denoising Autoen-coders
multi-target detection and tracking
