Unlock instant, AI-driven research and patent intelligence for your innovation.

Classification loss-based generative adversarial network restored abnormal vein image identification method and device

A vein image and identification method technology, applied in biometric identification, biological neural network model, character and pattern recognition, etc., can solve the problems of poor repairing effect of vein image, fracture, blurred vein texture and so on.

Active Publication Date: 2021-02-12
北京圣点云信息技术有限公司
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0007] The technical problem to be solved by the present invention is to provide a classification loss-based generative adversarial network repair abnormal vein image recognition method and device to solve the problem that traditional repair algorithms do not have a good repair effect on vein images, and vein texture structures are prone to blurring and fracture etc., resulting in poor recognition performance of the vein recognition system for abnormal vein images

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Classification loss-based generative adversarial network restored abnormal vein image identification method and device
  • Classification loss-based generative adversarial network restored abnormal vein image identification method and device
  • Classification loss-based generative adversarial network restored abnormal vein image identification method and device

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0079] With reference to accompanying drawing 1 shown, a kind of recognition method that the present invention relates to based on the generation confrontation network repairing abnormal vein image of classification loss comprises the following steps:

[0080] 1) Perform normalization processing on several defective vein images A and normal vein images B corresponding to several defective vein images one-to-one to form a training set. In this embodiment, since there are almost no suitable label images for abnormal vein images in reality, it is impossible to use real abnormal vein images to construct a training set. Therefore, 200,000 normal vein images B are selected and multiplied by any mask image M (the damaged part is 0, and the undamaged part is 1) to obtain 200,000 damaged vein images A, and the normalized size of the vein image is 160* 64, and use this as the training set.

[0081] 2) Build a generator network based on SK-ResNet, and add the SK-ResNet structure to the ...

Embodiment 2

[0118] As shown in FIG. 6 , this embodiment relates to a classification loss-based recognition device for generating an adversarial network to repair abnormal vein images, which includes:

[0119] 1) A normalization processing module, which is used to perform normalization processing on several defective vein images A and normal vein images B corresponding to several defective vein images one-to-one to form a training set. The normalization processing module is used to realize the embodiment 1 The function of step 1);

[0120] 2) The generator network building module is used to build a generator network based on SK-ResNet. The SK-ResNet structure is added to the encoder of the generator. The SK-ResNet structure is an embedded SKNet module added to the ResNet residual module. module, the generator network building module is used to realize the function of step 2) of embodiment 1;

[0121] 3) The discriminator network building module is used to build a CNN-based discriminator n...

experiment example

[0133] In order to verify the repair effect and recognition performance of the method of the present invention on abnormal vein images, 50 finger images of users were selected, 10 finger images for each user, including a normal image library and a damaged image library (mirror dirty block) , a total of 50*10*2=1000 pairs, the image size is 500*200, and the normalized image size is 160*64.

[0134] The damaged images in the damaged image library are repaired by the method generator repair module of the present invention and the traditional Crinimisi method respectively, Fig. 7 a) is the finger vein image of the real dirty block, and Fig. 7 b) is the finger vein image repaired by the present invention , Figure 7c) The finger vein image repaired by the traditional Crinimisi repair method, it can be clearly seen that although the traditional Crinimisi method has repaired the dirty block, there are still some block effects, and the gray scale of the finger vein image has faults and ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention relates to a classification loss-based generative adversarial network restored abnormal vein image identification method and device. The vein image identification method comprises the following steps of 1) forming a training set; 2) establishing a generator network based on the SK-ResNet; 3) establishing a discriminator network based on the CNN; 4) establishing a classification network based on the SK-ResNet; 5) training the classification network based on the SK-ResNet by adopting the normal vein image B; 6) repairing the defective vein image A to form a repaired vein image C; 7) updating generator parameters; 8) updating parameters of the discriminator network based on the CNN; 9) carrying out iterative training for a plurality of times; 10) obtaining a restored vein imagebased on the optimal generator model; 11) extracting vein features, and 12) performing identification. According to the method, the SK-ResNet structures are added into the generator network and the classifier network, so that the parameter quantity of the model is greatly reduced, the gradient disappearance problem of the convolutional neural network is relieved, and the vein image feature extraction capability of the network is improved.

Description

technical field [0001] The invention belongs to the field of image processing and biological feature recognition, and in particular relates to a recognition method and device for repairing abnormal vein images based on a classification loss-based generative confrontation network. Background technique [0002] Vein recognition includes palm vein recognition, hand vein recognition, etc. As an emerging identification technology, its market is expanding and has broad prospects because of its strong anti-counterfeiting ability, rapid recognition and high accuracy. Dirty blocks on the mirror surface of the vein image imaging equipment, skin peeling on the user’s fingers and palms, etc. will cause some features of the vein image to be missing, forming an abnormal vein image, making it difficult to extract the vein features of this part, and ultimately affecting the recognition of the vein recognition system. performance. Therefore, how to improve the vein recognition system has im...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G06K9/00G06K9/62G06N3/04
CPCG06V40/10G06V40/14G06N3/045G06F18/214G06F18/24
Inventor 赵国栋李学双张烜
Owner 北京圣点云信息技术有限公司