Development Of Binocular Pupil Detection System And Its Algorithm Research

Pupillary light reaction is a kind of autonomic nervous stress reaction,and its objective detection index can be widely used in the early discovery of diseases,diagnosis differential,monitoring and evaluation of postoperative condition for optic nerve diseases.At present,manual inspection methods are used in clinic in the domestic and overseas and have great subjectivity.Therefore,it is of great significance to use objective and quantitative index instead of qualitative detection method for the early diagnosis,clinical monitoring and prognosis of optic nerve diseases.In view of the above requirements,this paper designed a set of binocular detection system for quantitative detection of the relative afferent pupillary defect(RAPD).Firstly,according to the overall requirements of binocular detection system,the hardware and software modules were designed.The hardware system includes the selection of light sources,the acquisition of images,the electrical system and the structure of the system.The software system includes the user information management module,the light source control module,the image acquisition module,the image processing and analysis module and so on.Secondly,the image processing algorithms involved in pupil detection were researched,including the preprocessing methods like image smoothing,binarization,morphological methods and the pupil segmentation methods like edge detection,Hough circle detection,contour tracking and so on.According to the characteristics of the pupil images,the advantages and disadvantages of each method were analyzed.Finally,in order to process the images acquired by the binocular detection system designed by this paper,the appropriate preprocessing methods were selected.And an improved contour tracking algorithm was proposed to finally realize the real-time pupil detection under photic stimulation.Experimental results show that the system proposed by this paper has good accuracy and robustness,and can meet the detection speed of 300 frames per second.