基于融合运动特征和深度学习的电厂人员行为识别

Behavior recognition of power plant personnel based on fusion of motion features and deep learning

电厂人员行为的准确识别与分析,对于维护电厂安全运行具有重要意义。文中使用融合运动特征的深度学习算法,建立了一套电厂工作人员的行为识别系统框架。为了提高动作识别精度、解决人体骨骼识别问题,通过引入图卷积网络建立多层时间空间融合的图卷积运算人体行为识别模型。针对单一网络检测准确率与鲁棒性低的问题,文中基于传感器网络建立了无线通信信道模型和电源管理方案,并得出视频传输优化方案,从而改善实际系统的通信质量。从实验验证结果可以发现,将所提算法搭载于多传感器网络中,对于电厂工作人员不安全行为的识别准确率可达90%以上,具有良好的工程应用前景。

Accurate identification and analysis of power plant personnel behavior is of great significance for maintaining the safe operation of power plant.In this paper,a deep learning algorithm based on motion feature fusion is used to establish a set of behavior recognition system framework for power plant workers.In order to improve the accuracy of action recognition and solve the problem of human skeleton recognition,the graph convolution network is introduced to establish a multi-layer time space fusion human behavior recognition model.Aiming at the problem of low detection accuracy and robustness of single network,the wireless communication channel model and power management scheme are established based on sensor network,and the video transmission optimization scheme is obtained,which improves the communication quality of the actual system.From the experimental results,it can be found that the recognition accuracy of the proposed algorithm in the multi-sensor network can reach more than90%,which has a good engineering application prospect.