目前用來做為個人身份辨識的生物特徵，已有瞳孔、容貌、指紋等，其中利用指紋來辨識是最常為人所使用的方法之一。它除了擁有辨識所需的唯一性以及不變性之特質外，在價格與效果的整體考量上，比採用其他的生物特徵划算許多，而且在法律的認證上，指紋辨識也佔有重要的地位。所以如何正確地分辨一枚指紋與其他指紋是否相同，在個人身份辨識上是一個很重要的課題。 從指紋擷取器所獲得的指紋影像，因按捺壓力大小不同以及機器所產生的雜訊，使得指紋影像的品質不如我們所預期地好，造成辨識錯誤的情況發生。本文先將指紋影像做前處理，試圖去增強及還原原始的指紋影像，並提出一個快速的指紋特徵搜尋法，去找出指紋特徵位置座標，再以粒子群最佳化演算法(PSO)，去與資料庫中的指紋做比對，從模擬結果來看，本文之演算法能更正確及快速地辨識個人身份。最後，將所提出之方法，在可程式化之系統晶片(SOPC)上，以嵌入式軟核心處理器來實現，能做到更快速且即時(On-Line)的指紋辨識系統。 Biometric features, such as pupils, faces and fingerprints, have been used to develop the personal identification system at present. Among all of them, the fingerprint identification is one of the most commonly used methods. It has not only the characters of the uniqueness and the time-invariance when recognizing but also lower costs and higher performance while comparing with other biological features. And the fingerprint identification also plays an important role in the legal authentication. Therefore, it is quite an important issue to distinguish accurately fingerprints from the others on the personal identification. The quality of the fingerprint image obtained from a sensor is not good as we expect because of the different pressure and the noise caused by the sensor. And it leads to the wrong identification result. In this paper, we first make a series of image pre-processes in order to enhance and restore the original fingerprint image. And then we propose a fast minutiae extraction method to find the coordinates of minutiae. After the minutiae extraction, we use a particle swarm optimization method (PSO) to match the fingerprints in a database. The experimental result shows that the personal identification is more accurate and faster via our algorithm. Finally, we apply our proposal to a system on a programmable chip (SOPC) by using the soft core embedded processor. And we can get a faster and on-line fingerprint identification system.
