The intravenous drug dispensing robot needs to handle as many as hundreds of drugs and hundreds of different types of drugs. The identification and measurement of different types of medicine bottles is the key to precise dispensing. In this paper, aiming at different tasks of intravenous drug dispensing robot, such as the classification of drug bottle types and the measurement of key dimensions of drug bottles, a machine vision method and deep learning framework YoloV5s are adopted to realize an efficient and stable drug bottle identification and measurement method. Combined with the measurement accuracy of the machine vision method and the insensitivity of the deep learning method to the background and light source, the accurate identification and measurement of the medicine bottle in the different dispensing process are realized. Finally, the effects of background and ambient light sources on the recognition results are quantitatively tested through experiments.