| With the application scope of information secure field, the biology recognize and net safe are developing rapidly. Finger recognize of biology recognize is more important. However, the fingerprint recognize technology is based on living fingers. So, how to distinguish live fingers is an important problem.The fingers can be forged, but the character and texture of the fake fingers are different from living fingers. Some are coarser while some are slicker, these are small different from live ones. Based on these points, the principal component analysis is applied to the characters, wavelets is used for the very little texture.The method of this paper is to get many statistical characters of fingers first, including first order character:energy, entropy, image mean, variance.Based on texture analysis, calculate co-occurrence matrix so that to get second order characters: CS. and CP. They are seen as properties of fingers with orient float, ridge mean, median and variance. The principal component analysis is applied to the many characters so to reduce dimension. More than 80% information two dimension vectors of origin fingers are used to label a finger swatch. The first dimension (the first principal component) vector is X-axis, the second (the second principal component)is Y-axis, they are used to get cluster center.Secondly,two level wavelets decompose are used, then reconstruct image with six details coefficients and proper threshold value,the difference of the later image and origin image is another character of the fingers. So we get a three dimension vector with the first two characters to label a finger swatch. Character vector range is calculated from tests to distinguish living fingers depending on distance between the finger and vector.125 living fingers, and 106 fake fingers are'used in this test and correct verification rate is 86.79% using texture analysis and the principal component analysis, correct verification rate is 96.42% if wavelets analysis is added. |
PDF Full Text Request