| Campus is a densely populated place with a large number of potential safety hazards.It is very important to accurately judge the actions of personnel and then analyze the risks to protect the safety of personnel.In fact,human motion recognition is a kind of multi-classification problem,which plays a crucial role in intelligent monitoring application system.A monitoring system capable of stable operation in today's complex daily environment has become an urgent need.In order to create a safe campus environment,this paper proposes a comprehensive human motion recognition method based on convolutional neural network,the traditional monitoring system can not timely analyze human behavior.This method can identify human behavior in real time.Experiments have proved this method has good accuracy and robustness,it can be applied to intelligent video surveillance system.The work of this paper is divided into three stages:(1)the human body target detection stage,which mainly applies the mainstream,low hardware cost,and realizes the simple ViBe algorithm and Otus+Sobel operator(OSobel)combined method(OSViBe)for human body target detection.The experiment proves that this method can quickly and effectively obtain the complete human body target area.(2)the feature extraction stage,this stage mainly designs GTNet extract features from the preliminary pre-processed images in the previous stage,and generates network based on regions to screen effective ROI areas;(3)the action classification stage,the GT classification module was used to classify the precise ROI selected in the RPN stage,and finally the classification of human movements was achieved,and the relevant data sets were used for training and testing.Finally,the classification results of the OSViBe+GTNet proposed in this paper are compared with the results of other advanced classification models.The experimental evaluation data show that the proposed method is more accurate and has certain practical application value. |
PDF Full Text Request