Research Of Dorsal Hand Vein Recognition Based On Bit-Plane

In the era of rapid development and application of the Internet+economic model,recognition under weak constraints such as cross-time,cross-equipment and cross-geography is an inevitable demand and trend in biometrics technology.Under the weak conditions,the collected dorsal hand vein images have some differences in brightness,rotation,translation,these differences lead to a lower recognition rate.Therefore,how to improve the robustness of dorsal hand vein recognition under weak constraints requires constant research and verification.The main work and innovation of this paper are summarized as follows:(1)Researched and analyzed the data characteristics of different types of dorsal hand vein images(gray and binary images),the image which preserved the veins was divided into 8 bits plane according to the coding principle of the image.Take the multi-angle rotation integral as the core,determined the best combination of experiments by experiment with different types of image separately.Experiments showed that the third bit-plane and multi-angle rotation integration can make the recognition rate of dorsal hand vein of single device up to 90.4%.(2)Aiming at the angle and scale transformation problem of the collected dorsal hand vein images,a method of dorsal hand vein recognition based on multi-angle rotating integration combine with discrete cosine transform was proposed.The bit-plane image is continuously rotated under a certain angular interval and was integrated to get the stacked vectors,an optimal feature matrix for classification identification is intercepted in the low frequency region after DCT,a descriptor which described the characteristic of the dorsal hand vein texture with a feature vector of 400 dimensions was obtained.The selected rotating angle 15 degrees,the size of the best feature matrix intercepted 20×20 and the third bit-plane can make the recognition rate of dorsal hand vein of single device up to 99.5%,the recognition rate of dorsal hand vein of cross-device up to 84.8%.(3)For the problem that dorsal hand vein recognition of cross-device was not ideal,a method of dorsal hand vein recognition based on block mutual information of bit-plane was designed.Divided the bit-plane into block,determined the best size of block,compared the three different mutual information calculation modes,a new descriptor which described the characteristic of the dorsal hand vein texture with a feature vector of 2592 dimensions was obtained.The best block size selected 20×20,eight-neighborhood of mutual information computing model and the third bit-plane can make the recognition rate of dorsal hand vein of single device up to 99.5%,the recognition rate of dorsal-hand vein of cross device up to 90.2%.Through the abovementioned work,recognition rate of single-device exceeds 99.5%,recognition rate of dorsal hand vein of cross-device from 82.60%up to 90.2%,the work of the paper was effectively verified.