Abstract: Human is the basic unit of manufacturing resources in the workshops.Due to the high degree of uncertainty in human's autonomous activities,it has always been difficult to obtain the information about the location and quantity of the staffs in the digital twin workshops.Concerning the urgent needs of real productions,an adaptive workshop staff recognition network (namely,Adaptive Rec-network) was proposed,which used the human in the loop approach to enhance the detection in the complex environment in the workshop.Comparing with the our former 3-Stage Cascade Convolutional Neural Network (3-Stage CCNN),the experiments demonstrated that the proposed model was more efficient,accurate and adaptive,and provided stable and accurate data for the digital twin.

Key words: workshop staff management, digital twin model, adaptive learning, visual enhanced detection

摘要： 作为车间制造资源的基本组成单位,人员主观活动具有高度不确定性,数字孪生车间中人员位置和分布信息的获取一直是难点。从现实生产生活的迫切需求出发,构建了一种自适应车间人员识别网络,并基于人在回路思想对车间复杂环境的检测效果进行增强。实验证明,所提方法与已有的三阶段级联卷积神经网络相比,检测效率和准确率更高、自适应性更强,可以为数字孪生模型提供稳定准确的支撑。

关键词: 人员智能管控, 数字孪生模型, 自适应学习, 视觉增强检测