The Design Of Dynamic Pupillary Light Reflex Detector

The detection and identification of human eyes and pupil is an important researchtopic of modern computer technology, it involves many disciplinary such as medical,artificial intelligence, pattern recognition, computer vision and other interdisciplinaryresearch scope. Pupillary light reflex (PLR) was proved to be a effective method toevaluate visual function for retinitis pigmentosa (RP) or other patients with low vision.The traditional examination methods for the human eye and pupil used by most of themedical institutions at home and abroad mainly based on visual electrophysiologicalexamination, include electro-oculogram (EOG), visual evoked potential (VEP) andelectro-retinogram (ERG), etc. These traditional detection methods are difficult to detectthe pupillary light reflex quantitatively and accurately, because theymay bringdiscomfort to participants or only work with the aid of psychophysics inspection orintuitive qualitative observation.Based on the existing theory of detection of the eye and pupil, in the specified lightstimulation conditions, this paper proposed an improved method to detect and recognizeeye and pupil from the video image sequence. The main contents are as follows:It’s necessary to design the light stimulator according to the detecting environment,to provide several different stimulation conditions with different colors, light intensitylevels, time stimulating. The light stimulator was mainly made of MCU and stimulatinglight source. The MCU completed the independent operation of bottom hardware, andthe communication with the computer application to receive control parameters, andthen parsed the means to drive the light source. The stimulating light source wascomposed of red, blue and white LED, and was placed evenly according to the circularsymmetric layout, in order to produce uniform scattering in specified spaceenvironment.This paper put forward a real-time eye and pupil detection method under specifiedstimulation referred above. The preprocessing about filtering and image normalizationwas necessary. Image segmentation would benefit to extract interesting area, at the sametime to reduce amount of data processing. The image was partitioned into severaldifferent classes based on iterative segmentation model of change proportional of imageinformation distribution, gray and entropy. And then features were extracted fromsuspected target area, by which the real target was classified out. At last, the parameters of location, area and radius of pupil were extracted and calibrated to the GUI.According to system requirements, dynamic tracking and detection strategy wasput forward. Once upon eye and pupil was detected, it’s better to track the target in thevideo image sequence continuously subject to accuracy and real-time performance.Once it did not detect the target continuously, it’s asked to detect the pupil again inaccordance with the update strategy.The application software with video acquisition, command control and imageprocessing was designed, at the same time humanized operation interface and datastorage were necessary too.The proposed pupillary light reflex detecting system can detect the situation andchange of pupil accurately under different light conditions, providing strong support inthe relevant field. The experimental results show that the proposed system has goodrobustness and accuracy.