Four-value weight and multiple classification-based human face feature extraction method

Abstract

The invention discloses a four-value weight and multiple classification-based human face feature extraction method. The method comprises the steps of constructing a human face training sample database; establishing a convolutional neural network; adjusting a caffe framework; preprocessing a human face image sample, inputting the human face image sample to the convolutional neural network for performing training, until the network is completely converged, and storing a generated human face identification model; and preprocessing a to-be-extracted human face image, producing a mean value file, inputting the mean value file to the human face identification model to obtain a feature graph, rotating the feature graph for multiple different angles to extract features respectively, performing addition fusion on different angle features of the same image, and finally obtaining a main human face feature. The method has the beneficial effects that the problems of huge memory consumption and insufficient storage space of network training is radically solved; and the feature with a strongest expression capability is obtained in a multi-feature extraction fusion mode, so that the human face identification accuracy is remarkably improved.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a face feature extraction method based on four-valued weights and multiple classifications. Background technique [0002] Face recognition technology is constantly iteratively updated. In this process, face recognition is more and more widely used in various fields. What comes with it is that not only high recognition rate is required for face recognition, but also high efficiency. The most urgent problem is to use the model on the mobile terminal to quickly and accurately identify the target. Based on this problem, if a fast and accurate identification method can be developed with a very small model on the mobile terminal, it will have a great impact on applications such as access control and attendance systems, social security, public security, justice, criminal investigation, and personal information security. s help. [0003] Deep learning is used to study the face...

AI Technical Summary

Problems solved by technology

The first way to train the network with large data sets is effective. Many methods now use this method to train the network to obtain better test accuracy, but collecting large data sets is a relatively difficult task. At the same time, the network Putting a large data set in the network will inevitably lead to slower network operation and larger models, which is not conducive to application on small devices; the most intuitive way to deepen and widen the network is to increase parameters, increase model size, and reduce computing speed. Reduce the efficiency of feature extraction; extracting high-dimensional features for similarity comparison will also reduce the efficiency of feature extraction and increase the size of the model. The low-dimensional features of information can better meet the needs of use

For many networks currently used for face recognition, a classification loss function is used for classification learning. A softmax function can classify faces, but the learned features are not too distinguishable, and it is very likely that different people Face recognition as the same person, or face recognition of the same person as different people

Images