Real-time print localization on pharmaceutical capsules for automatic visual inspection

The goal of this work was to enable real-time visual inspection of printed pharmaceutical capsules, as the information printed on the capsules is crucial for unambiguous identification of active substances and dosages contained in the capsule. We propose a template matching technique to localize and detect print on capsules. For this purpose, large image distortions related to capsule shape were first eliminated by estimating the principal axes of the capsule and transforming the capsule image to cylindrical coordinates. Then, a fast and efficient implementation of normalized cross correlation was used to localize the printed image or text for which the template was selected a priori. A dedicated database of printed capsules was used to assess and confirm the feasibility of the proposed method in terms of accuracy (<;0.5 pixels), reliability (100%) and speed (<;2 ms). In our case, the pixel size was 70 μm, therefore, localization accuracy of up to 0.5 pixels is sufficient for subsequent image analysis and defect classification, since defects of size less than 70 μm are hard to detect visually, which additionally justifies the feasibility of the proposed method for the real-time automatic visual inspection of printed pharmaceutical capsules.