本研究旨在建立一個指紋辨識系統驗證平台，可分別針對指紋影像品質、指紋特徵點擷取正確性和指紋比對性能等功能模組進行獨立評估。此一驗證平台分為指紋資料庫管理、指紋影像品質評估、指紋特徵點擷取評估和指紋比對性能評估四個子系統。我們結合了網頁前後端技術，並使用Docker建立多個容器來實作這個平台軟體。最後採用多個實驗指紋資料庫來驗證這個平台的功能性。實驗結果表明，本研究能夠對指紋辨識三個功能模組進行個別評估，指紋影像品質模組能夠評估不同感測器或取像環境所獲得的影像品質指標；指紋特徵點擷取模組能夠計算指紋特徵點擷取的正確性，並可編輯和標註錯誤特徵；指紋比對模組產出代表辨識系統性能的相似度表和FAR、FRR不同閾值下的趨勢圖與ROC曲線。由於採用MIAT方法論設計，使得本平台的系統架構更加模組化；使用Docker容器讓本平台適用於跨平台的開發和佈署。;This study aims to establish a verification platform for fingerprint recognition systems, which can independently evaluate modules such as fingerprint image quality, accuracy of fingerprint feature point extraction, and fingerprint matching performance. The verification platform consists of four subsystems: fingerprint database management, fingerprint image quality assessment, fingerprint feature point extraction evaluation, and fingerprint matching performance evaluation. We have combined web front-end and back-end technologies and utilized Docker to create multiple containers for implementing this platform software. Finally, multiple experimental fingerprint databases were used to verify the functionality of this platform. The experimental results demonstrate that our study enables the individual evaluation of the three functional modules of fingerprint recognition. The fingerprint image quality module can assess image quality indicators obtained from different sensors or imaging environments. The fingerprint feature point extraction module can calculate the accuracy of fingerprint feature point extraction and allows editing and labeling of erroneous features. The fingerprint matching module produces similarity tables, and trend charts representing the performance of the recognition system of FAR and FRR at different similarity threshold values, as well as ROC curves. By adopting the MIAT methodology, the system architecture of this platform becomes more modular, and the use of Docker containers enables cross-platform development and deployment of the platform.
