Research And Implementation Of Dsp-based Embedded Fingerprint Identification System

The traditional high performance Automated Fingerprint Identification System(AFIS) is mostly based on personal computer. However, it is not enough for portability and convenience, and it is more and more distinct that the AFIS tends to be a miniaturized and embedded system. Considering that reason, the fingerprint identification technology , correlative digital image algorithm and a set of algorithms for embedded system of automated fingerprint identification, which mainly includes fingerprint image pre-processing, feature extraction and feature matching, are studied.In the aspect of fingerprint image pre-processing, firstly, we propose an adaptive normalized algorithm, which melts adaptive threshold segmentation algorithm, to achieve the normalized and segmentation of the input image. Secondly, we adopt the enhanced algorithm based on Gabor filter, which includes calculation block oriented image, frequence image, Gabor filter , Gausi filter and other spacial filters. Finally, we complete the binarization and thinning of the enhanced image.In the aspect of fingerprint image feature extraction, firstly, all minutiae are extracted from the pre-processed image by using feature extraction templates. Secondly, bogus minutiae are deleted, the true minutiae and its ridges information sampled are saved by using ridge tracking technique so that they can be used in the matching of fingerprint images.In the aspect of fingerprint image matching, firstly, we adopt the information about ridge of feature points to reduce the work for selecting reference points, and then we transform the minutiae positions into polar coordinates to deal with the displacement between template image and input image. Finally, we accomplish fingerprint image matching adapting bounding box to effectively deal with the nonlinear distortion of fingerprint image.All algorithms are tested in the PC and DSP system. Experimental results show that the pre-processing algorithm can improve the quality of the fingerprint image efficiently, the feature extraction algorithm extracts the true minutiae and deletes the bogus minutiae exactly, and the matching algorithm effectively solves the coordinates alignment problem and the nonlinear distortion problem.