abstract |
The invention discloses an identity verification method based on an unsteady iris video stream bionic neural network. The method includes the following steps: in step 1, 1000-2000 images are collected; in step 2, the grayscale images of the iris are converted into 256×32 dimensional images. Template iris normalization enhanced image; step 3, perform convolution processing on all template iris normalized enhanced images; step 4, extract iris feature data of 32-bit template iris normalized enhanced image; step 5, obtain the template of the template tester Class recognition of iris categories; step 6, intercept 90 iris images as test iris images; step 7, convert into enhanced images of 256×32 dimensions; step 8, convert into 4 groups of 30×254 dimensional two-dimensional arrays; step Nine, extracting iris feature data; step ten, obtaining an identity verification result. Beneficial effect: The setting of the video stream also avoids the influence of the non-steady iris on the feature expression, which is beneficial to the setting and adjustment of the iris feature when the number of training irises is small. |