Image block weighted convolutional neural network-based face recognition method

Abstract

The invention discloses an image block weighted convolutional neural network-based face recognition method. The method includes the following steps that: a sample database is constructed, edge detection is performed on a sample picture, and a face contour picture is intercepted; five portions in the face contour image are positioned, segmentation is performed according to positioning, so that local pictures can be obtained, and the gray scale variance mean of the local pictures is calculated; the face contour picture and six local pictures, which belong to the same sample picture, are together put into the training of a convolutional neural network; and segmented pictures of a pictures to be recognized are put into the trained convolutional neural network, so that a recognition result can be obtained. According to the image block weighted convolutional neural network-based face recognition method of the invention, both local characteristics and global characteristics are considered, so that a system can have a better recognition effect. According to the image block weighted convolutional neural network-based face recognition method of the invention, the image block weighted convolutional neural network is adopted, and thus, compared with a traditional face recognition method, the image block weighted convolutional neural network-based face recognition method of the present invention can improve the recognition rate of face recognition.

Description

technical field [0001] The invention relates to the technical field of face recognition, in particular to a face recognition method based on a convolutional neural network weighted by image blocks. Background technique [0002] With the development of computer technology and Internet technology, face recognition has also become a popular research object in the field of computer vision in recent years. There are broad application prospects. Convolutional neural network is developed on the basis of traditional neural network. It is an efficient recognition method. In recent years, it has been greatly developed and gradually applied to various fields. Convolutional neural network can extract detailed structural information from input images. , and at the same time, it can make these structural information have spatial invariance such as depth rotation, which is very suitable for detection and recognition problems in images. [0003] The current mainstream method is to use con...

AI Technical Summary

Problems solved by technology

But most of the time the training is based on the whole image, and usually many people will use uniform block, but the uniform block will not be able to segment the same area of ​​the human face due to the relationship between angle and posture, so it cannot effectively extract the image Local information, so the recognition rate will also be limited

Images