虚拟视频眼动分析设计

发布时间：2018-01-19 09:14

本文关键词： 视线追踪眼动分析头部运动数据融合　出处：《西安工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：视线追踪技术在人机界面交互、助残、疲劳驾驶检验、航空医学研究等领域有广泛的应用前景。头部运动是人视线注视过程中的重要的伴随行为,对于人类视觉注意有着重要的意义。本文针对眼动分析技术的研究,在面部识别和特征分析的基础上,首先改进视线追踪方法和头部运动跟踪法:对于视线跟踪,本文采用非穿戴式宽视野红外相机获取人面部图像,通过Adaboost算法实现面部识别,然后采用ASM算法对获取的面部图像进行实时特征跟踪,并截取眼球区域图像,然后通过霍夫变换算法和图像阈值处理获取并跟踪眼球区域图像中普尔钦斑和瞳孔中心,最后采用改进的四光源透视法,建立四光源普尔钦斑与瞳孔关系模型,通过多项式估计获得视线注视点。对于头部运动跟踪,本文同时采用基于惯性传感器和基于POSIT摄像机观测法的头部姿态空间运动的跟踪。首先本文通过四阶龙格库塔法、毕卡逼近和加速度插值积分,实现惯性传感器的姿态运动跟踪。然后通过POSIT迭代算法对图像进行解算获得头部运动信息。最后结合两种方法的特点,通过数据融合,取长补短,有效提高头部运动跟踪测量效率。最后,本文结合软件体系结构,实现了头部运动与视线追踪数据融合测量系统。系统由多个软件构件构成,输入数据通过连接件传入构件,输出所需要的数据,构成一套具备完整功能的软件。
[Abstract]:Eye tracking technology has a wide application prospect in human-machine interface interaction, disability, fatigue driving test, aeronautical medical research and other fields. Head movement is an important accompanying behavior in the process of human gaze. It is of great significance to human visual attention. This paper focuses on the research of eye movement analysis based on facial recognition and feature analysis. First of all, improve the method of line of sight tracking and head motion tracking: for line of sight tracking, this paper uses a non-wearable wide-field infrared camera to obtain human face images, and realizes face recognition through Adaboost algorithm. Then the ASM algorithm is used to track the real time features of the acquired facial image and to capture the eyeball region image. Then through the Hough transform algorithm and image threshold processing to obtain and track the Purchin spot and pupil center in the eye region image. Finally, the improved four light source perspective method is used to establish the relationship model between the four light source Purchin spot and the pupil. The point of view fixation is obtained by polynomial estimation. At the same time, this paper uses the inertial sensor and POSIT camera observation method to track the head attitude space motion. Firstly, the fourth order Runge-Kutta method, Bika approximation and acceleration interpolation integral are used in this paper. The attitude tracking of the inertial sensor is realized. Then the head motion information is obtained by the POSIT iterative algorithm. Finally, combining the characteristics of the two methods, through the data fusion, learn from each other. Finally, this paper combines the software architecture to realize the head motion and line of sight tracking data fusion measurement system. The system is composed of several software components. The input data is transmitted to the component through the connector, and the required data is outputted to form a set of software with complete function.
【学位授予单位】：西安工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】