指紋特徵點擷取的精確度是影響指紋辨識的關鍵因素。本研究提出一個模組化的指紋特徵點擷取的軟體架構，利用MIAT方法論以及軟體高階合成的方法，將特徵點擷取的軟體設計成一個模組化的架構，使得不同指紋特徵擷取演算法都可以應用此架構來評估與測試。本研究使用啟發式方法、SIFT演算法與基於卷積神經網路的FingerNet，三種不同的方法來驗證此一軟體架構的可調性。實驗表明，使用者可以透過參數來切換不同的演算法，由於軟體模組使用正規化的輸入輸出格式，使得不同演算法可以在相同的軟體架構上進行精確性的性能評估。本研究成果有利於針對不同的指紋特性、不同取像環境、不同感測器特性，快速評估適用的指紋特徵點擷取演算法。;The accuracy of fingerprint feature extraction is a critical factor affecting fingerprint recognition. This study proposes a modular software architecture for fingerprint feature extraction using the MIAT methodology and high-level software modeling and synthesis techniques. By using designed modular software framework, it allows different fingerprint feature extraction algorithms to be evaluated and tested within this architecture. Three distinct methods, Heuristic, Scale-Invariant Feature Transform (SIFT), and FingerNet, were used to verify the adaptability of the software architecture. Experimental results shows that users can switch between different algorithms by using different arguments of the software. The standardized input-output format enables accurate performance evaluation of different algorithms within the same software architecture. The findings of this study provide a convenient way to evaluate suitable fingerprint feature extraction algorithms for different fingerprint characteristics, imaging environments, and sensor properties.
