Research On Small Area Fingerprint Identification Technology Based On Android System

Fingerprint recognition technology has been widely used in mobile terminal devices such as smart phones and tablet PCs.However,as mobile terminal devices are moving toward thinner and lighter,capacitive fingerprint collectors are becoming smaller and smaller,resulting in an area of fingerprint images collected.The smaller.Small-area fingerprint images contain fewer detailed feature points(endpoints and bifurcation points).Traditional f-ingerprint recognition methods based on detailed feature points will result in correct acceptance rate reduction and false acceptance rate increase in small-area fingerprint recognition.The high phenomenon can not meet the application requirements of mobile terminal equipment,so small-area fingerprint recognition technology has become one of the research hotspots in the field of fingerprint identification.In order to improve the correct acceptance rate of small-area fingerprints,this paper has carried out in-depth research on small-area fingerprint recognition technology from image processing and deep learning.The main work and innovations are as follows:1)A small-area fingerprint recognition method based on binary descriptor is proposed.Firstly,the Gaussian difference method is used to detect the key points,and centered on key points,extracting the binary descriptor consisting of the ultrashort binary descriptor of the fingerprint original image and the detailed information of the fingerprint binary image.Secondly,when the key point pair is matched,the Hamming distance is used to realize rough matching and fine matching of detailed information of ultrashort binary descriptors and binary images respectively.Finally,in order to reduce the fingerprint template storage space and improve the fingerprint recognition rate,a fingerprint feature splicing strategy based on the binary descriptor is realized.2)A small area fingerprint identification method based on densely connected convolutional networks is proposed.Firstly,the image enhancement is used to the fingerprint image.And the deep learning model for fingerprint feature extraction is constructed,which makes full use of the advantages of feature reuse of densely connected convolution network to realize small area fingerprint recognition.3)The fingerprint recognition function based on Android system is realized.According to the framework design of the Android system,the fingerprint driver design of the Linux kernel layer,the function interface defined in the hardware abstraction layer,and the encapsulation of the small area fingerprint recognition program based on the binary descriptor are orderly implemented.And The capacitive fingerprint collector SW9251 and the Hikey960 development board were selected to verify the fingerprint recognition function based on the Android system.In order to verify the effectiveness of the two methods proposed in this paper,9600 fingerprint images were collected by the capacitive fingerprint collector SW9251 as a self-built database.The above two methods respectively obtain the correct recognition rate of 99.46%and 97.47%in the self-built database,which is higher than the traditional fingerprint recognition method based on the detailed feature points and basically meets the application requirements of the mobile terminal device.