基于高光谱成像技术的运动人脸姿态识别方法  

Moving face gesture recognition method based on hyper-spectrum imaging technology

作　　者：芦鹏[1] LU Peng(Criminal Investigation Police University of China public Security Information Technology and Intelligence College,Shenyang 100854,China)

机构地区：[1]中国刑事警察学院公安信息技术与情报学院,沈阳100854

出　　处：《激光杂志》2020年第11期105-109,共5页Laser Journal

基　　金：国家社科基金(No.16BTQ064);辽宁省自然科学基金项目(No.20180550046);中国刑警学院2019重大项目培育计划资助项目(No.D2019001)。

摘　　要：运动人脸姿态受到不规则光谱分布的干扰导致检测的准确性不好,提出基于高光谱成像技术的运动人脸姿态识别方法。采用高光谱视觉成像技术进行运动人脸姿态图像视觉信息采样,采用多层特征分解方法提取运动人脸姿态高光谱图像的边缘轮廓特征量,结合视觉信息重构的方法进行运动人脸姿态光谱特征检测,建立运动人脸姿态的光谱分布结构模型,采用活动轮廓套索模型进行运动人脸姿态高光谱图像角点信息标记,根据角点特征提取结果进行人脸的关键特征点标注,根据运动人脸姿态高光谱图像的特征分布实现运动人脸姿态识别。仿真结果表明,采用该方法进行运动人脸姿态识别的准确性较高,特征分辨能力较强,提高了运动人脸姿态信息检测和特征提取能力。The accuracy of moving face gesture detection is not good due to interference of irregular spectral distribution.A moving face gesture recognition method based on hyper-spectral imaging technology is proposed.The hyperspectral vision imaging technology is used to sample the visual information of moving face gesture image,and the edge profile feature quantity of moving face gesture hyper-spectral image is extracted by multi-layer feature decomposition method.The spectral feature detection of moving face gesture is carried out with the method of visual information reconstruction,and the spectral distribution structure model of moving face gesture is established.The active contour lasso model is used to mark the corner information of moving face gesture hyper-spectral image,the key feature points of the face are marked according to the results of corner feature extraction,and the moving face gesture recognition is realized according to the feature distribution of moving face gesture hyper-spectral image.The simulation results show that the proposed method has high accuracy and strong feature resolution ability,which improves the ability of moving face gesture information detection and feature extraction.

关 键 词：高光谱成像 运动 人脸 姿态 识别 特征提取 

分 类 号：TN247[电子电信—物理电子学]