Machine Vision Based Novel Scheme for Largely, Reducing Printing Errors in Medical Package

Aiming at the problem of misprint or obscure of production date, batch number and the validity on the medicine package, a delay-based misplaced difference scheme for medical information detection is put forward. To be specific, medical images with delay in packaging are acquired by a vision detection system, based on which, character images are obtained through misplaced subtraction operation; and a convolution kernel is designed based on gray value distribution of the character images for multi-step convolution to remove speckle noise. Then a specific operation with corrosion and dilation is further utilized to remove speckle noise and enhance the target character area. In the end, the modified weighted median filter is adopted for noise inhibition to further improve the recognition accuracy. Experimental results show that when the double image overlap η is 80% and threshold λ (the percentage of non-zero gray values in each convolution block) is 60%, the scheme’s recognition can be as accurate as 97.8% and detection speed can reach up to 0.1373 s/image, the detection precision and efficiency can satisfy medical information recognition requirements in medical package.