The work aims to research the recognition of drug images based on machine vision technology, and seek better algorithms and the appropriate external conditions in the process of drug recognition. Through theoretical analysis combined with HALCON software, a series of experiments were carried out on drug recognition, and the effects of illumination system and image processing algorithms on recognition results were analyzed.The research showed that, on the basis of successful recognition, the recognition difficulty was obviously related to the working environment, and various processing algorithms, target thresholds and system complexity could have a significant impact on the recognition results.Compared with the ordinary median filtering algorithm, the processing results of improved median filtering algorithm are significantly improved. By adjusting the external light source and background reference, it is helpful to improve the processing quality. Through the reasonable combination of RGB and HSV threshold processing algorithms, the recognition result of the system can be significantly improved. The research results can provide important reference for drug assembly and related industrial intelligence.