Face spoofing detection method and system based on color channel difference image features

Abstract

The invention discloses a face spoofing detection method and system based on color channel difference image features. The method comprises the following steps: selecting a face image after preprocessing video data; extracting the face image to obtain a plurality of color channel images; calculating a difference value between the color channel images to obtain a plurality of color channel difference value images, and performing normalization processing to obtain a normalized color channel difference value image; adding the set attention module into a deep convolutional neural network to construct an attention network; combining the cross entropy loss and the verification loss into a total loss through weighting; inputting the normalized color channel difference graph into an attention network for noise feature learning, updating a network weight coefficient according to a loss value, training the attention network, and storing a model and a weight of the attention network after trainingis completed; and predicting a classification result by the trained attention network. According to the method and system, the spoofing noise characteristics can be accurately extracted, and the generalization ability of the face spoofing detection model is improved.

Description

Technical field [0001] The invention relates to the technical field of face detection and recognition, in particular to a face deception detection method and system based on the characteristics of a color channel difference map. Background technique [0002] Face spoofing detection is to determine whether the face recognition process is a real face or a static or dynamic fake face. These fake faces may usually be printed face photos, replayed face videos or 3D masks, etc. The existing face spoofing detection technology One method is to implement detection based on traditional manual features, for example, by extracting texture features, and the other is to perform detection based on deep learning; while methods based on manual feature extraction are easily affected by lighting conditions and scenes, and The extracted features are not rich enough to meet the requirements of detection accuracy; most methods based on deep learning directly use RGB or grayscale images as network inpu...

AI Technical Summary

Problems solved by technology

[0002] Face spoofing detection is to judge whether it is a real face or a static or dynamic fake face in the process of face recognition. These fake faces may usually be printed face photos, replayed Face video or 3D mask, etc., one way of the existing face deception detection technology is to implement detection based on traditional manual features, for example, by extracting texture features, etc. The other way is to detect based on deep learning; and based on The method of manually extracting features is easily affected by lighting conditions and scenes, and the extracted features are not rich enough to meet the requirements of detection accuracy; most methods based on deep learning directly use RGB or grayscale images as the network for training. In this way, the features learned by the network may not be effective features for distinguishing live images, but fit the features of face recognition. Under the supervised training of "true / false" labels, the model is easy to only Pay attention to the difference of non-deceptive information in the two types of training data (such as differences in face structure, background content, etc.), so the model has high detection accuracy in the database, but the cross-database performance is greatly reduced, and the generalization is poor

Images