摘要
为了对大型施工设备的生产率进行准确的实时监测和分析,提出了一种基于计算机视觉的设备识别和生产率监控系统。该方法利用Faster-RNN对视频序列中的挖掘机和卡车进行识别检测,然后通过ASFormer对挖掘机动作进行动作分割。基于行为分割的结果对挖掘机的工作效率进行计算。比较了人工计算与行为分割网络的准确度的差距,对2台不同的挖掘机进行动作的识别和分析,结果显示平均准确度分别为91.7%和94.8%。因此,当多个场景并行处理时,可以大大节省人力成本,验证了本方法的有效性和实用性。本研究为慧工地的数字化管理提供了有效的技术基础,开辟了计算机视觉应用的新场景,提出了一种新的计算挖掘机生产效率的计算因子,更适合于通过统计挖掘机动作来进行效率计算。
In order to accurately monitor and analyze the productivity of large-scale construction equipment in real time,a method of equipment identification and productivity analysis based on computer vision is proposed.The method consists of two stages:firstly,the excavator and truck in the video sequence are identified and detected by Faster-RNN,and then the excavator action is segmented by ASFormer.The work efficiency of excavators is calculated based on the results of recognition and behavior segmentation.For two different excavators,the average accuracy is 91.7%and 94.8%respectively.When multiple scenarios are processed in parallel,the labor cost can be greatly saved,which verifies the effectiveness and practicability of this method.Therefore,this study provides an effective technical basis for the digital management of hui site,opens up a new scene of computer vision application,and puts forward a new fast calculation of excavator production efficiency,which can realize the real-time calculation of the work efficiency of construction equipment.It can play an important role in improving the production efficiency of construction site equipment and ensuring the safety of staff.
作者
陈雪健
秦水介
白忠臣
郭媛君
杨之乐
CHEN Xuejian;QIN Shuijie;BAI Zhongchen;GUO Yuanjun;YANG Zhile(College of Big Data and Information Engineering,Guizhou University,Guiyang 550025,China;Guizhou Key Laboratory for Photoelectron Technology and Application,Guizhou University,Guiyang 550025,China;Shenzhen Institute of Advanced Technology,Chinese Academy of Sciences,Shenzhen Guangdong 518055,China)
出处
《智能计算机与应用》
2023年第4期111-116,121,共7页
Intelligent Computer and Applications
基金
国家自然科学基金-地区项目(61865002,62065002)。
关键词
深度学习
机器视觉
行为分割
特征提取
视频检测
生产率检测
监控系统
deep learning
computer vision
action segmentation
feature extraction
video processing
productivity monitoring
monitoring system