基于低配置DSP内核的指纹识别软件设计

Design of Software for Fingerprint Recognition Based on Low-configure DSP

低配置DSP内核指纹识别系统上应用的算法不同于普通的PC运行环境下运行的算法,它需要在算法软件以及和硬件相关软件上采取不同的方法。该文针对低配置DSP内核下运行指纹识别算法要求低内存和低速的特点提出相应的设计方案。其中的增强算法采用了简化的方向场和方向滤波的方法,以及采用局部统计的二值化方法和形态处理的细化算法。细节点提取则采用模板提取和后处理来提高性能。匹配方法则采用二次比对结合平移旋转配准来实现。算法在DSP应用中根据低资源的实际情况,采用基于窗口的处理方法和各种汇编级别的优化,实验结果表明整个算法和软件的设计的效果是不错的。

The algorithm applied in fingerprint recognition system of low - configure DSP core is different from the algorithm applied in ordinary PC. This paper aims at the low memory and low speed of fingerprint recognition system in low configure DSP situation and presents the corresponding design. The algorithm of enhancement adopts a way of simplified orientation field calculation and orientation filter. Binarization using local statistical way and the thinning algorithm of mathematical morphology are adopted. The minutiae extraction adopts model extraction and post - procession to improve efficiency. Matching method combines the translation and rotation alignment with two - time match. The whole algorithm applied in DSP has an other improvement based on the low resource. It adopts the procession based on windows and is optimized with all kinds of methods in assembly level. The experiment result shows that the effect of design of the whole algorithm and software is good.