摘要
为了改善电力服务行业场景复杂,服务行为识别困难的问题,提出了一种供电营业厅服务行为识别融合网络。该网络主要包括时空分割网络模型和改进C3D网络模型。首先,在从视频中提取光流帧和RGB帧。其次,将提取出的光流帧和RGB帧带入时空分割网络和改进C3D网络经过训练,从而有效提取动作特征和图像特征。最后,在分类层,计算每个网络对每类服务动作的识别准确率,通过Softmax公式确定权重,并得到最终动作识别结果。仿真阶段,以南方电网公司提供的服务视频数据集为例,对所提模型进行验证。仿真结果表明,所提方法识别准确率为98.99%,召回率为90.2%,F分数为94.39%。仿真结果进一步验证了所提模型对服务行为具有较高的准确性和稳定的识别率。
In order to solve the problem of complex scenarios and difficult service behavior recognition in power service industry,a service behavior recognition fusion network for power supply business hall is proposed.The network mainly includes spatiotemporal partition network model and improved C3D network model.Firstly,optical flow frames and RGB frames are extracted from video.Secondly,the extracted optical flow frames and RGB frames are brought into the spatiotemporal segmentation network and the improved C3D network for training,so as to effectively extract action features and image features.Finally,in the classification layer,the recognition accuracy of each network for each type of service action is calculated,the weight is determined by Softmax formula,and the final action recognition result is obtained.In the simulation phase,the service video data set provided by China Southern Power Grid Corporation is taken as an example to verify the proposed model.The simulation results show that the recognition accuracy of the proposed method is 98.99%,the recall rate is 90.2%,and the f score is 94.39%.The simulation results further verify that the proposed model has high accuracy and stable recognition rate for service behavior.
作者
衡星辰
陈英达
付彦哲
HENG Xingchen;CHEN Yingda;FU Yanzhe(China Southern Power Grid Co.,Ltd.,Guangzhou,510700)
出处
《微型电脑应用》
2024年第1期145-148,共4页
Microcomputer Applications
关键词
电力系统
营业厅
行为识别
光流
时空分割
power system
business hall
behavior identification
optical flow
spatiotemporal segmentation
