摘要
针对施工现场环境复杂,难以高效管理的问题.提出了基于工地场景的深度学习目标跟踪算法,辅助施工顺利进行.根据工地现场目标的连续性,构建增强群跟踪器,提升目标成功跟踪的概率.然后从滑动窗口、Stacked Denoising Auto Encoder(SDAE)和Support Vector Machine(SVM)三方面组建深度检测器.在滑动窗口方面:从梯度角度建立模型实现窗口自适应.在SDAE算法方面:构建反向算法微调网络参数.优化SVM算法降低跟踪时目标漂移和跟踪失败的概率,最终实现目标高精度跟踪.通过实验表明本文提出的算法可有效对目标进行跟踪,实现动态管理.
Construction site is difficult to be effectively managed owing to its complex environment.A deep learning target tracking algorithm based on construction site scene is proposed to assist the construction progress.Firstly,according to the continuity of the target in the site scene,the enhanced group tracker is constructed to improve the successful probability of target tracking.Then,the depth detector is constructed with sliding window,stacked denoising auto encoder(SDAE)and support vector machine(SVM).Sliding window:a model is built from the gradient angle to realize window adaption.SDAE algorithm:the reverse algorithm is built to fine-tune network parameters.Optimized SVM algorithm reduces the probability of target drift and tracking failure.Finally,high precision tracking is achieved.Experiments show that the proposed algorithm can track the target effectively and realize dynamic management.
作者
马少雄
邱实
唐颖
张晓
MA Shao-xiong;QIU Shi;TANG Ying;ZHANG Xiao(Xi’an University of Technology,Xi’an,Shaanxi 710048,China;Shaanxi Railway Institute,Weinan,Shaanxi 714000,China;Key Laboratory of Spectral Imaging Technology CAS,Xi’an Institute of Optics and Precision Mechanics,Chinese Academy of Sciences,Xi’an,Shaanxi 710119,China;Chengdu University of Technology,Chengdu,Sichuan 610059,China;Northwest University,Xi’an,Shaanxi 710127,China)
出处
《电子学报》
EI
CAS
CSCD
北大核心
2020年第9期1665-1671,共7页
Acta Electronica Sinica
基金
中国科学院“西部之光”项目(No.XAB2016B23)。
