RGB激光扫描的面部特征点识别研究

传统方法在进行危险预警时,需要被测者穿戴整套仪器设备,成本高,稳定性不好,为此,提出一种基于激光扫描的多源脸部信息感知危险预警方法。在进行激光扫描前,通过直接标定法对相机进行标定,并且提出适于RGB激光扫描的颜色标定法,令采集的点云数据中所有点均和一组RGB颜色数据相应,从而得到脸部彩色三维重建结果。利用激光扫描仪得到测角与测距,获取测点的三维坐标,和扫描点构成人脸三维点云数据图,对点云数据进行分析处理,得到多源人脸信息。依据眼睛特征点与嘴部特征点判断人体是否处于疲劳状态,从而完成危险预警。实验结果表明,所提方法危险预警准确性高。

Recognition of facial feature points by RGB Laser Scanning

In the traditional method of hazard early warning,the subject needs to wear a whole set of instruments and equipment,which has a high cost and lousy stability.Therefore,a multi-source face information perception hazard early warning method based on laser scanning is proposed.Before the laser scanning,the camera is calibrated by the direct calibration method,and a colour calibration method suitable for RGB Laser scanning is proposed,which makes all the points in the collected corona data correspond to a group of RGB colour data,to obtain the 3D reconstruction results of face colour.Using a laser scanner to get angle measurement and distance measurement,obtain the three-dimensional coordinates of measurement points and scan points to form the human face’s three-dimensional point cloud data map.Point cloud data is analysed and processed to get multi-source face information.According to the eye feature points and mouth feature points,we can judge whether the human body is fatigued,to complete the danger warning.The experimental results show that the proposed method is of high accuracy.