Development of a Genetic Method for the Recognition of Medical Packaging

The problem of automation of definition of medical packages on arrival at a drugstore is considered. The main methods of pattern recognition in information systems are analyzed. Problems in the field of recognition and methods of their solution are considered. To automate the process of sorting only incoming drugs to the pharmacy department, a modification of the pre-prepared neural network for the robot sorter was developed, which on the basis of this neural network will give the exact name after packaging analysis, which will reduce the time to fill out the invoice. Image augmentation was performed, which allowed to improve the result. The introduction of a genetic algorithm for finding and optimizing the neural network architecture was also carried out. As a result, an increase in recognition accuracy was obtained. This solution will simplify the work for the first table and pay more attention to customers to avoid queues.