Under the trend of intelligent manufacturing and flexible manufacturing, human-machine collaboration has been the mainstream in production. Therefore, as a key factor affecting the efficiency of human-machine collaboration, it is very important to identify and analyze human actions in production line. However, cameras are often unable to get close to the operators considering the complex environment of industrial sites, while the identification accuracy reduces greatly because multi-person wearing the same uniform in industrial sites at the same time. Therefore, this paper proposes an automatic multi-person detection and pose estimation method based on object detection and pose estimation framework, which not only effectively distinguishes the identity information of the multiple operators, but also provides accurate body joint features for identifying the actions of multiple target operators.