Tablet defect detection during pharmaceutical bottling process is extremely crucial to ensure product safety, quality, and regulatory compliance. Traditional defect detection methods include human inspection, which is less efficient, time-consuming, and high labor cost. Though AI-powered defect detection driven by machine learning algorithms may offer unmatched speed and precision, their complex arithmetic and high cost may not be the best option in industry. To resolve this dilemma, a real-time tablet defect detection model using center gradient variation algorithm is developed in this study. First, the image sensor combined with convex lens is used to scan the tablet shape. Second, the analog OS signal generated from the image scanning is binarizedinto a series digital code array. Third, the center point coordinate from each scanned line is positioned. Fourth, a slew rate limiter is introduced to remove unwanted noise from the center point form line. Finally, the mean gradient variation ( MGV ) is established and used to evaluate the defect level. Experimental results verify that the proposed algorithm can identify the defect status up to 100% accuracy either minor or severe defect.
Lin et al. (Thu,) studied this question.