3D Fixation Point Detection Based On 3D Reconstruction Technology And Its Visualization Method

The perception and reconstruction of 3D scene has always been an important issue in computer vision and the basis of human-computer interaction in real environments.This article discusses the design architecture and 3D reconstruction system of the visual inertial odometry in detail.A visual inertial odometry technology is used to realize a more stable SLAM system,and an indoor scene reconstruction method based on RGB-D data is proposed.Analyze the potential deficiencies and propose possible improvements.Eye-tracking is an important means of human-computer interaction.Conventional eye-tracking technology is mainly aimed at gaze estimation in 2D scenes,which cannot meet the needs of human interaction with real 3D environments.This article summarizes and analyzes 3D fixation point detection methods for head-mounted eye trackers.Discuss the potential error sources and technical difficulties in these methods,and provide clear guidelines for designing lightweight 3D gaze estimation systems.On this basis,a 3D gaze estimation method is proposed.This method directly maps the 2D pupil position in the eye camera to the 3D gaze direction without the need for 3D observation from the eye camera.In this thesis,a system for estimating the 3D fixation point in the world coordinate system is designed by fusing the gaze estimation of the head-mounted eye tracker,the head posture estimation of the visual inertial odometry and the environmental point cloud data of the depth camera.Compared with the previous method,this method can accurately estimate in world coordinates and is more robust to environmental changes.Visualization is an important tool for analyzing eye movement behavior.As a visualization method,the heat map can very intuitively express the area that the observer is gazing at,but often ignores the saccade information that expresses the shift of the fixation point.Based on the conventional heat map generation method,this thesis proposes a new dynamic heat map generation method based on eyetracking data.This method uses a mixed data structure of fixation points and saccades,and adds a heat map deformation method for saccade data.The heat map generated by the method in this thesis adds the function of indicating the path of the gaze area transfer while maintaining the original function of displaying the gaze area.