Research On Human Behavior Recognition Method In Production Process Based On Depth Vision

With the innovation and integration of the new generation of information technology and manufacturing industry,informatiozation and intelligence have become an important trend in the development of the manufacturing industry.The management of manpower,machine,material,method and environment in the production processes have become an important strategy to achieve intelligent manufacturing.Compared with other production factors,workers are highly autonomous,which makes it a chore to manage their behavior in the production process effectively.In this paper,we study the working activities in production process,recognize abnormal behaviors,and standardize the production activities,which is of great significance to improve the management of the workshops.This paper proposes a working activity recognition approach based on human skeleton information,which is aimed at the activity recognition and monitoring of workers in production workshops.The 3D depth vision sensor is utilized to collect the body skeleton joint position data,and a standardized reconstruction methods is used for the pre-processing of skeleton joint data.Specifically,the posture recognition model based on transient features and the action recognition model based on sequent temporal features are then constructed separately.For the posture based on transient features,the posture expression of personnel based on the angledistance composite feature of skeleton vector is constructed.Then,logistic regression algorithm,decision tree algorithm and support vector machine algorithm are used to realize the gesture recognition of personnel based on transient features.For the human action based on features of sequence in time,based on the spatial-temporal feature RGB image of synthetic personnel action,a deep convolutional neural network model is constructed to realize the production action recognition in space-time domain.Finally,the software and hardware system for personnel production behavior data collection and behavior recognition was developed and built based on the research of human behavior recognition,and the collection of personnel behavior data as well as the effective behavior recognition were realized in the laboratory environment.