Research On Information Extraction Of Gaze Area Of Scene Target In Gaze Tracking

With the development and gradual improvement of computer vision,the application of object detection and gaze tracking technology has gradually become widespread,and has gradually penetrated into every aspect of people's lives.Its importance has also become increasingly prominent,and it has begun to occupy great applications in many fields,such as human Machine interaction,fatigue detection,air traffic,etc.This article is based on the research on the monitoring and early warning of the operator.The effect of the task performed by the operator is critically affected by the control panel.In order to execute the control panel more safely,the operator needs to monitor its real-time status,and Responses to abnormally displayed conditions should be made in a timely manner.This article mainly studies the information extraction of the target's gaze area in the gaze tracking,combines the gaze tracking technology to calculate the position of the operator's gaze,and uses the convolutional neural network to extract the information of the objective gaze area that exists objectively.Based on the matching between the point of sight of the operator and the dashboard information in the control panel,the research content of this paper is as follows.First,we need to extract the line-of-sight parameters of the operator.After detecting the human eye area,we can extract the line-of-sight parameters.The line-of-sight parameters include the spot center coordinates and the pupil center coordinates.Thresholding is used to obtain a binary map of the human eye area.The binary map is filtered and the Huff circle transform is used to calculate the pupil center coordinates.Then,in the process of gaze estimation,this paper uses a spatial geometric model based on cross-ratio invariance proposed by Yoo and Chung to calculate the gaze point.This method does not require an initial calibration operation,and at the same time can maintain high accuracy and has good adaptability to head movements.Finally,the operator's line of sight is matched with the dashboard's characteristic information,and the pixel area corresponding to the operator's line of sight on the dashboard is analyzed.Based on computer simulation technology,use the computer screen to simulate the control panel area,match the identified and segmented dashboard area with the operator's line of sight,and use the Mask R-CNN network structure to detect the dashboard and extract valid instruments.Characteristic information,when the point of sight of the operator does not match the pixel area corresponding to the dashboard,some responses will be made,such as voice reminders.This article combines gaze tracking and convolutional neural networks to monitor and alert the operator's physical and mental state,in order to improve pilot selection accuracy and training efficiency and reliability.After the corresponding simulated flight training test,the method proposed in this paper has achieved good results in training,which has great reference value for China's aviation industry and improves the safety and reliability of the aviation industry.The effective implementation of this paper also provides new theoretical methods and technologies for the evaluation,selection,monitoring and early warning of operators and research in related fields.