Implementation Of Real-time Eye Tracking System Based On GPU

As an important equipment for diagnosing diseases such as inner ear disease,sports disease,Meniere's syndrome,and benign paroxysmal positional vertigo,the video nystagmography developed rapidly.The current video evoked electrokigograph on the market has a higher complexity of the pupil tracking algorithm because of the accuracy requirement.This complex method of pupil center positioning requires extremely high computer hardware requirements and greatly increases the cost of the entire set of equipment.Due to the use of a more complicated pupil center tracking algorithm,the video evoked electrosurgery has poor real-time handling of the pupil center.Therefore,the video nystoelectric imager has not been used at large in the hospital at this stage.In particular,the video evoked electrocardiograph of small and medium-sized hospitals is an extremely rare diagnostic instrument.Based on the knowledge of image tracking and positioning,this paper combines GPU parallel computing to develop a set of video-nystagmography apparatus with accurate tracking,strong real-time performance and reasonable cost.Utilizing GPU parallel computing to optimize the image tracking and positioning program and a tracking-accurate but complicated pupil tracking positioning algorithm,the real-time tracking and positioning of the pupil center is accomplished on the basis of ensuring the accuracy of pupil center tracking.This article utilizes the GPU's powerful parallel computing capabilities to support the medical disciplines of the information sciences,improve the existing video evoked electrocardiograph on the market,expand the scope of use of video evoked electrograms,and implement video evoked electrograms.The real-time nature of the instrument allows the video nystagmography to be widely used in the research of hospitals and universities.The main work of this article includes:1)Using the common image processing methods in mathematical morphology to process the pupil images,a new pupil location tracking algorithm is used to locate and track the pre-processed pupil images to obtain the diagnosis of the pathogenesis of certain vestibular diseases.Shock signal.The eye tracking system has high tracking accuracy and reasonable cost.2)Utilize the GPU's powerful computing capabilities to optimize the image processing program in the mathematical morphology and the proposed pupil tracking and positioning program.Including the optimization of image segmentation,enhancement,median filter and erosion dilation in image preprocessing,and optimization of pupil tracking algorithm by using parallel protocol,so that the eye tracking system under GPU architecture can process more efficiently.The whole eye tracking system is more real-time.3)Build an experimental platform to compare the effects of the eye tracking system under the GPU architecture and the eye tracking system under the CPU architecture in the same experimental platform,including three from the program processing efficiency,pupil positioning accuracy and hardware costs The analysis and comparison of experimental results are carried out.4)According to the analysis of experimental results,this paper summarizes the methods and steps for optimizing different programs using GPU parallel computing from the optimization of the GPU parallel computing program for the pupil tracking program.