Brand identification of roller‐type date stamps based on impression features

Roller-type date stamps have become increasingly used in recent years within Chinese low-cost manufacturing and packaging sectors to mark production and expiration dates on food, pharmaceuticals, and other small consumer goods. These inexpensive devices are also employed in administrative and commercial documentation, such as contracts, invoices, and logistics records. Given their widespread use in production and documentation, roller-type stamp impressions are encountered in forensic casework involving document forgery, counterfeit labeling, and product traceability disputes. This study presents a quantitative framework for the brand identification of roller-type Chinese date stamps through morphological feature extraction and machine learning analysis. Sixteen commercial models were systematically evaluated under 36 simulated stamping conditions, yielding 17,280 impression samples. Seventeen morphological features were extracted from the fixed Chinese characters "" (year), "" (month), and "" (day) and refined through interval overlap rate (IOR) analysis and principal component analysis (PCA). Four supervised classifiers-Gradient Boosting Decision Tree (GBDT), Random Forest (RF), Support Vector Machine (SVM), and Naïve Bayes (NB)-were trained and validated to assess model robustness and discriminative ability. GBDT and RF demonstrated superior accuracies (96.76% and 95.66%) in laboratory tests and maintained strong field performance (89.06% and 90.88%). The results confirm that roller-type stamp impressions exhibit quantifiable, brand-specific morphological characteristics that can be reliably recognized using data-driven models. The proposed approach offers a rapid, non-destructive, and scientifically robust tool for forensic brand identification and evidential authentication.