Based On Wavelet Transform Is Fast Fingerprint Identification System Design And Implementation

Fingerprint identification technology in daily life is widely used in areas such as security, and its application is more for embedded system, the embedded system is the key research objects in the field of science and technology, so the fingerprint identification algorithm research are also gradually thorough, this thesis deeply studied a fingerprint image fast recognition algorithm, and successfully implemented on JZ2440- ARM development board of the fingerprint image acquisition and fingerprint identification.At the beginning of the fingerprint image recognition, we need to make the fingerprint image segmentation and gray variance segmentation algorithm has been widely used in fingerprint image segmentation.However, the choice of threshold has an important impact in the performnce of the gray variance segmentation algorithm.In this paper,we present a adaptive threshold updating algorithm and the image area is divided into 8*8 small pieces to calculate the variance and mean of the small area.When the small area meets the conditions, we will substitude the variance and mean into the update formula threshold to update it. Adaptive threshold updating algorithm we propose can automatically adjust the threshold depending on the fingerprint image and we can distinguish the foreground region or the background region better. In order to improve the operation speed of the algorithm, this paper presents a determining the center square algorithm for fingerprint image which is based on direction of three-level correction that the direction angle of the next block is corrected by using the direction angle of the upper block.The algorithm is able to remove the artifacts caused by fuzzy, fracture, missing blocks and other phenomena.The algorithm is also able to remove the local characteristics of the fingerprint image caused by uneven force in the fingerprint collection.Then, the fingerprint image is transformed from spatial domain to frequency domain by wavelet transform. The multi-scale analysis is performed in frequency domain to extract the texture information of the fingerprint image, and improve the accuracy of fingerprint identification.Fingerprint acquisition part of the hardware circuit is designed in this paper, we using semiconductor fingerprint sensor MBF200 in the image acquisition chip.In order to transfer the datas between MBF200 and development board,we use the USB bus interface mode.In the part of the software we use the Linux operating system and its transplantation.We use QT creator integrated development environment to design a fingerprint image applications to match the window and achieve fingerprint acquisition and calibration. Finally, through the implementation of the embedded system, the result will show that the proposed fingerprint recognition algorithm can effectively identify the fingerprint image.