Identification of various defects in pharmaceutical tablets using image processing techniques

Abstract

Objective: Our aim is to identify the damaged tablets from the manufacturing line using image processing techniques and remove them before packaging. Methods: The various problems posed during inspection are broken tablets, corner chips, black or other color spots in tablets, empty blisters (without one or more tablets or capsules), foreign particles/color variation in the tablets/capsules, improper sealing, etc., Image processing techniques will be used for defect detection. Results: Tablets are available in packed forms that are usually transparent, semi-transparent or opaque. Euclidean distance was employed for detecting defects, during testing that had a similarity of 100 for tablets with no defects, for defective blisters had similarity ranging from 98 to 41. Empty blisters had a similarity of 0 on comparing with trained images. Conclusion: Similarity measuring based technique can accurately detect defects in the pharmaceutical tablets, hence can be adopted for removing such blisters from the manufacturing line itself.