Study On Cognitive Characteristics Of Human Vision On Object Movement In Driving Context

Object dynamic tracking has important research implications in areas such as intelligent surveillance,video object tracking,and scene classification,which involves the ability of the human visual system to rapidly localize attention to certain types of objects in complex scenes that may contain individual variations of global topological features or spatial variations of global motion features.These two global features have not yet been fully investigated in terms of the cognitive properties of visual attention.In this paper,we focus on the two global features of "topological features" and "motion state features" to study the cognitive properties of visual attention.And based on the experimental study of cognitive properties of "virtual 2D graphical cognitive",we focus on exploring the cognitive properties of visual attention in the process of motion target recognition in the experiment of "virtual 3D driving cognitive".The main results of the research are as follows:1.The first part of the work proposes a new experimental method of EEG analysis for exploring the property of human vision on attention cognition.The method system is based on the Unity3 D virtual reality development platform to build "virtual 2D graphical cognitive" and "virtual 3D driving cognitive" experimental scenarios,which are presented in the HTC Vive Pro Eye head-mounted display,and using Logitech G29 secondary development technology and EEG interface technology were used to control the driving of the car in the scene.The Ag/AgCI electrode caps were used to monitor the real-time EEG response data of the participants during the experiment.The results of the ERP component wave assessment of the EEG data can show that the N1 component wave amplitude of ERP is more susceptible to the motion characteristics of the stimulated subjects and varies significantly.2.The second part of the work proposes a new experimental method of Eye-tracking analysis for exploring the property of human vision on attention cognition.The method uses response surface analysis techniques to rationalize the experimental variables based on optimizing the experimental scenarios constructed in the above work,and uses the Tobii Eye-tracking instrument built into the VR glasses to collect human eye gaze data.The evaluation of the Eye-tracking dataset showed that the global motion feature on the spatial extent of the target object preferentially caused attentional capture during visual perception,with this feature accounting for 94.68% of the total dataset and its first gaze data accounting for 75.89% of the total first gaze dataset.3.The third part of the work proposes a new experimental method of EEG and Eye-tracking analysis for exploring the property of human vision on attention cognition.The method combines the relevant research techniques and data collection methods employed in the first two methods based on the optimization of the constructed experimental scenarios,and increases the sample size of the experimental data.The evaluation results of 3082 total first gaze datasets show that the global motion features of the target object are more preferentially perceived by visual attention,and the effect of high level of motion features on the dynamic capture of visual attention reaches highly significant.In summary,this paper explores the attentional cognitive properties of "topological features" and "motion state features" during visual perception by designing novel "virtual 2D graphical cognitive" and "virtual 3D driving cognitive" attentional features experiments.The results show that the global motion features of the object in space are captured more preferentially by the visual attention of human eyes than the global topological features of the target object.This conclusion is an extended development of the topological nature perception theory,which can provide an interpretable understanding of the currently prevalent visual attention phenomena,which will provide a new fundamental theoretical reference for a credible machine vision computational model.