Feature modeling method for human face identification

Abstract

The invention discloses a feature modeling method for human face identification. The method comprises the steps of presetting 22 key feature points, wherein the 22 key feature points specifically include two angular points of each eyebrow, two angular points of each eye, an uppermost point and a lowermost point of each eyelid, a nasal tip point, two nasal wing points, two angular points of the mouth, an uppermost point and a lowermost point of an upper lip, an uppermost point and a lowermost point of a lower lip, and a lower jaw point; according to a preset calibration sequence of the key feature points, manually calibrating the key feature points in a training sample; according to a human face image as the training sample, generating a group of feature point coordinate data, and forming a group of shape vector training samples; and according to the shape vector training samples, building a global shape model and a local texture model. A changing track of the positions of the selected key feature points can represent change of facial emotions more accurately, so that the accuracy of human face emotion identification can be effectively improved.

Description

technical field [0001] The invention relates to biometric feature recognition, in particular to a feature modeling method in face recognition. Background technique [0002] Face recognition technology generally includes four components, which are face image acquisition, face image preprocessing, face image feature extraction, and matching and recognition. Specifically: [0003] Face image acquisition and detection refers to the acquisition of video or image data including human faces through video image acquisition devices such as camera lenses, which can be static images, dynamic images, different positions, and different expressions of the acquisition object. [0004] Face image preprocessing refers to determining the part of the face from the collected image data, and performing image preprocessing such as grayscale correction and noise filtering, so that the subsequent face image feature extraction process can be more accurate and efficient. [0005] Face image feature ...

AI Technical Summary

Problems solved by technology

[0008] The inventor found through research that the accuracy of recognition is low when performing emotion recognition through the selected key feature points in the prior art

Images