基于深度学习的机房人物重识别研究

Research on computer room character recognition based on deep learning

机房内的各类设施通常由管理员统一调度,对视频中管理人员的路径追踪至关重要。为此该文提出一种基于深度学习的机房人物重识别方法,以为后续的责任追查过程提供依据。该方法以残差神经网络ResNet-50作为特征提取网络,并使用三元损失函数使模型更适用于人物重识别任务。以贵州省某电网系统机房作为试验平台进行测试。结果表明,该方法的准确率与召回率均满足实际要求,具备一定的工程参考价值。

All kinds of facilities in the computer room are usually dispatched by the administrator,which is very important for the administrator's path tracking in the video.In order to provide the basis for the follow-up process of responsibility tracing,a new method based on deep learning is proposed.In this method,the residual neural network resnet-50 is used as the feature extraction network,and the ternary loss function is used to make the model more suitable for human re recognition task.A power grid system room in Guizhou Province is used as the test platform for testing.The results show that the accuracy and recall rate of the method meet the actual requirements,and have certain engineering reference value.