Maxout multi-convolution neural network fusion face recognition method and system

A convolutional neural network and neural network technology, applied in neural learning methods, biological neural network models, neural architectures, etc., can solve problems such as uneven classification of positive and negative samples, increased network training time, and different performance, so as to reduce positive The effect of uneven performance of negative samples, improvement of face recognition accuracy, and improvement of training speed

Active Publication Date: 2017-09-22
XIDIAN UNIV
View PDF6 Cites 36 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Although the rapid development of convolutional neural network has brought great breakthroughs to face recognition, there are still many problems to be solved and overcome in the face recognition technology based on a single convolutional neural network, such as lighting, background and posture. It has an impact on the final result; different network structures perform differently on the same test set. Some convolutional neural networks perform well on negative sample pairs (two images that do not belong to the same person), and some convolutional neural networks perform well on negative sample pairs (two images that do not belong to the same person). The neural network performs well on positive pairs (two images belonging to the same person)
[0004] To sum up, the existing face recognition method based on a single convolutional neural network, due to its uneven classification of positive and negative samples, leads to low face recognition test accuracy
However, the existing face recognition methods based on multi-convolutional neural network fusion mostly use simple serial fusion or linear fusion, which leads to an increase in network parameters, increases network training time, and easily causes network overfitting.

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
  • Maxout multi-convolution neural network fusion face recognition method and system
  • Maxout multi-convolution neural network fusion face recognition method and system
  • Maxout multi-convolution neural network fusion face recognition method and system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0030] The present invention is a kind of face recognition method based on Maxout multi-convolutional neural network fusion, see figure 1 , including the following steps:

[0031] Assume that the application environment of the face recognition method based on Maxout multi-convolutional neural network fusion proposed by the present invention is the entrance of a railway station, and the purpose is to identify the face image detected by the surveillance camera and the face image of the criminal suspect provided by the public security department , to achieve this, the implementation steps include:

[0032] Prepare training data:

[0033] (1a) Collect the face images detected by the surveillance camera at the entrance of the railway station, and use the collected face images as the training database D 0 .

[0034] (1b) From the training database D 0 Select an image, use the regression tree combination algorithm and affine transformation to perform face alignment correction on ...

Embodiment 2

[0042] The face recognition method based on Maxout multi-convolutional neural network fusion is the same as embodiment 1

[0043] Wherein the process of constructing fusion network based on Maxout in step (2) includes:

[0044] (2a) Build the first convolutional neural network, see figure 2 , figure 2 (a) is the first convolutional neural network structure diagram, which includes an input layer, 5 convolutional pooling layers, 2 fully connected layers, and an output layer in sequence according to the direction of data flow. figure 2 (b) is a schematic diagram of the convolutional pooling layer structure, where the convolutional pooling layer includes 2 convolutional layers with a convolution kernel size of 3×3, 2 Relu activation layers, and a Max pooling layer. In this example, two convolutional layers are connected alternately with two activation layers. The first is the convolutional layer; the next output is the Max pooling layer. The Max pooling layer can be replaced...

Embodiment 3

[0049] The face recognition method based on Maxout multi-convolutional neural network fusion is the same as embodiment 1-2, wherein the method utilizing Maxout in step (2c) fuses the convolutional neural network in step (2a) and step (2b) , obtain the fused convolutional neural network, comprising the following steps:

[0050] (2c1) Intercept the input layer, 5 convolutional pooling layers and the first fully connected layer in the first convolutional neural network constructed in step (2a) to obtain the subnetwork S 1 .

[0051] In this example from figure 2 In the first convolutional neural network structure shown in (a), all layer structures from the input layer to the first fully connected layer are sequentially selected according to the data flow direction to form a subnetwork S 1 .

[0052] (2c2) Intercept the input layer, 5 convolutional pooling layers and the first fully connected layer in the second convolutional neural network constructed in step (2b) to obtain t...

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 provides a Maxout multi-convolution neural network fusion face recognition method. The technical problem of unequal classification of positive and negative samples in a single network is solved. The method comprises the steps that a regression tree combination algorithm and affine transformation processing training data are used to acquire a database after alignment; face images in the aligned database are extracted to acquire a database simply with face area images; two networks are built, and a Maxout method is used to acquire a fused network; the database simply with face area images is used to train the fused convolution nerve network to acquire a trained network model; and a test image is pre-processed to test the trained network model. According to the invention, a Maxout module is used to fuse two or more subsystems, and then is successively connected with a full connection module and an output module to acquire a complete face recognition system; the classification sensitivity of positive and negative samples of the convolution neural network can be balanced; and the face recognition accuracy is improved.

Description

technical field [0001] The invention belongs to the technical field of image processing, relates to convolutional neural network and face recognition technology, specifically a face recognition method and system based on Maxout multi-convolutional neural network fusion, which can be used for video retrieval, dynamic monitoring, identity recognition, Intelligent buildings and other fields. Background technique [0002] As a natural attribute of human beings, human face is a biological feature with great differences and easy acquisition. Therefore, face recognition technology has received extensive attention and research. Face recognition specifically refers to an authentication technology that analyzes and judges face images through computers and related algorithms. Face recognition technology is widely used. For example, in the field of public security criminal investigation, fugitives can be arrested by following the face recognition system at airports and stations; face r...

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
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/00G06N3/04G06N3/08
CPCG06N3/08G06V40/161G06V40/168G06N3/045
Inventor 侯彪焦李成张华王爽马晶晶马文萍冯捷张小华
Owner XIDIAN UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products