Research on Filter Rod Counting System Based on Computer Vision Technology

Le résumé fourni par la source

To address the need for automated counting of cigarette filter rods, this study proposes a high-precision counting method based on image recognition. This method achieves precise automatic localization and counting of filter rods through Gaussian blurring, edge detection, morphological opening operations, watershed segmentation, centroid marking, and circular feature identification methods etc. In tests of hollow filter rod samples across different batches, the proposed method achieved a counting accuracy of 99.46%, with an average processing time of 2.74 seconds per tray, and low dispersion of recognition results. Comparative experiments demonstrate that its comprehensive performance is significantly superior to the row-column mean method (94.27% accuracy, 1.31 s/tray) and the inverted trapezoidal standard gauge-based method (99.61 % accuracy, 17.85 min/tray), exhibiting prominent advantages in generalizability and robustness. This study provides theoretical support and technical references for enhancing counting precision in industrial automation of filter rod quantity statistics.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.