Face recognition processing method, face recognition processing device and mobile terminal

Abstract

The invention provides a face recognition processing method and a face recognition processing device, wherein the method comprises the steps of determining attitude adaptation key points which correspond with a face image; performing convolution processing on the face image by means of a trained convolutional network, wherein the number of various layers in the network structure matches a processing resource of terminal equipment, and the parameter of the various layers has attitude robustness; performing attitude robustness processing on a first characteristic graph which is output from a first node in the convolutional network according to the attitude adaptation key points, generating a second characteristic graph with attitude robustness response, and transmitting the second characteristic graph to the convolutional network through a second node which is adjacent with the first node, thereby extracting face characteristics which correspond with the face image. The face recognition processing method, the face recognition processing device and the mobile terminal have advantages of realizing high processing speed of a small system, improving robustness and accuracy in face recognition, and improving processing efficiency and user experience.

Description

technical field [0001] The present application relates to the technical field of image processing, and in particular to a face recognition processing method, device and mobile terminal. Background technique [0002] With the development of portable terminal devices, many devices support the face recognition function. Taking mobile phones as an example: applications such as unlocking mobile phones through face recognition. [0003] The current face recognition method with the highest recognition accuracy is a deep convolutional network. Different from traditional face recognition methods, deep convolutional networks have more layers. This multi-level structure can better express complex patterns and distributions in data, and can extract more robust and discriminative face features. [0004] However, the amount of computation and memory consumption of the deep convolutional network is much higher than that of the traditional method, generally more than 10 times, so the face ...

AI Technical Summary

Problems solved by technology

However, limited by power consumption and cost, the computing power and storage space of small systems are not enough to carry the face recognition algorithm based on deep convolutional network, and the performance of face recognition is poor

Images