本篇論文首先應用一個方法擷取感興 區塊，此方法可打破限制，利用非固定式的裝置收集掌紋及掌背血管影像。接著我們將掌紋及背血管進行影像融合，以得到比單張影像更豐富且有用的特徵。此外，我們利用新的特徵擷取方法來獲得影像中的特徵資訊。最後分別以樣板比對以及支援向量機兩種分類技術來區分是否為本人。實驗結果得到98. 73%的準確率，這說明了我們所提出的方法是有效果且具可靠性 。;Biometric verification gradually plays an important role and highly demand for security systems. There are many biometric features including fingerprint, iris, hand geometry and facial image that can be used for biometric verification. However, the performance of traditional uni-model biometric systems can not meet the damand in providing satisfactory anti-spoofing capabilities. Multi-model biometric systems are then emerging with more satisfactory performance than uni-model biometric systems because multiple information can be acquired from different biometric characteristics. In this thesis, two physical biometric features including palm print and palm vein are utilized in our biometric verification system.
In our approach, we devise a method to extract the region of interest (ROI) which relieves the limitations constrained by traditional docking devices. Then, palm print and palm-dorsa vein images are fused to form a new image for providing richer and more useful information. Next, the features adopted for verification are extracted by using Histogram Iterative Thresholding (HIT). Finally, template matching and support vector machine (SVM) are employed as the classifiers for identity verification. Experimental result shows that 98.73% accuracy rate can be achieved by using our proposed approach. It demonstrates that our proposed approach is efficient and robust in the application of biometric verification.
