摘要
针对目前已有的嵌入式指纹识别系统往往采用手工提取,不能自动学习并提取识别所需的特征及识别正确率仍然不高的缺点,提出一种基于自动编码器(Auto Encoder)和LSSVM的指纹识别系统。首先,提出采用FPS200作为指纹传感器采集指纹数据,然后将采集的数据经过滤波和二值化等预处理,通过比较差异算法获得Auto Encoder中的权值和偏置等参数,从而得到训练好的Auto Encoder用于指纹图像特征提取。最后,将自动提取的特征进行训练和分类,将投票最多的分类作为指纹识别的结果。通过测试表明,系统能较精确地实现指纹识别,具有收敛速度快、正确识别率高和匹配时间短的优点。
The existing embedded fingerprint identification systems can not automatically remember and extract the features for fingerprint identification, thus leading to a low identification ratio. To solve these problems, a fingerprint identification system based on Auto Encoder and LSSVM was proposed. First, the FPS200 was used as a fingerprint sensor to obtain fingerprint data, which was then pre-managed by filters and binaryzation. Second, contrast difference algorithm was used to get the parameters of the Auto Encoder such as weight and bias. The trained Auto Encoder was further applied to collect fingerprint image features. Third, the automatically extracted features were input to the trained LSSVM to be further trained and classified, and those with the most voters were regarded as the identification results. The test indicates that the proposed method can precisely recognize fingerprints. Compared with other methods, it is faster in convergence rate,higher in identification rate, and more efficient in fingerprint matching.
出处
《中国测试》
CAS
北大核心
2015年第8期71-74,93,共5页
China Measurement & Test
基金
国家自然科学基金项目(61079022)
关键词
指纹
识别率
匹配
自动编码器
fingerprint
identification rate
match
auto encoder
