The work aims to propose a method for rapidly recognizing the state of the elevator hall door based on the YOLO model for the problem of the elevator hall door packing state recognition in the flexible production line robot of elevator hall door. The industrial camera was used to capture the container image and make a sample training set. Then the training set was input into the target recognition classification detection model, and iterative training was performed by adjusting the network structure parameters. After testing and verification, the recognition method proposed had a success rate of more than 99% for hall door state recognition, and the recognition speed was obviously superior to the traditional machine vision processing algorithm. The rapid recognition method for hall door packing state proposed can effectively solve the problems of low recognition efficiency and high misjudgment rate of traditional machine vision processing algorithms due to complex and variable illumination factors in industrial environment, and can meet the beat requirements of production system.