Computer Vision for Pokémon Collectible Trading Card Identification: A Systematic Literature Review

The Pokémon Trading Card Game (TCG) includes high-value assets, yet the cataloging of large collections is still largely manual and error-prone. This systematic literature review (SLR) synthesizes evidence on computer vision pipelines for card detection/segmentation, OCR-based metadata extraction, and recognition strategies suitable for expanding catalogs. Following Kitchenham and Charters (2007), searches were conducted in IEEE Xplore, and Scopus, complemented with backward and forward snowballing in Google Scholar. After screening and quality assessment, eight primary studies were selected. The synthesis indicates that modern single-stage deep detectors outperform geometric baselines for localization under noise and reflections, OCR benefits from adaptive binarization combined with fuzzy matching to improve identification reliability, and closed-set multiclass classification is inadequate for dynamic TCG inventories. The review consolidates design recommendations and highlights research opportunities in reflection mitigation (e.g., protective sleeves), open-set recognition, public benchmarks, and real-time on-device optimization.