Research On Spatial Location Of Human Eyes Pupils Based On Infrared Images In Monocular Vision

As the window of the soul,eyes are one of the main ways for human beings to understand and obtain information from the outside world.At present,global visual health problems are serious: there are more than 2.2 billion people with visual impairment or blindness,among which at least 1 billion people have visual impairment problems that could be prevented or need to be solved.For commonly used clinical ophthalmic disease detection instruments,the imaging optical axis of the instrument should be aligned with the central point of the pupil before detection,so as to ensure that the light emitted by the light source is incident into the eye of the tested person through the pupil.In practice,although the medical staff can realize the alignment through manual adjustment,manual alignment has certain requirements on the operator’s proficiency and the degree of cooperation of the tested,which is tedious and time-consuming.Therefore,automatic eye alignment is particularly important,and the premise of automatic eye alignment is to get the spatial position of the pupil.Therefore,this paper proposes a spatial positioning technology of human pupil in monocular vision based on infrared image,which can achieve the purpose of obtaining the spatial position coordinates of human pupil.The main research contents are as follows:(1)Face features,human eye features,pupil positioning needs and the difficulties of pupil positioning technology were deeply analyzed.The pupil spatial positioning technology was divided into pupil plane positioning technology and pupil single visual distance measurement technology,and the corresponding image acquisition platform was built.(2)A method of locating the pupil plane from coarse to fine is proposed.Firstly,in view of the large number of parameters and complex calculation problems in the extraction of human eye feature points in the traditional VGG13 network,the network model was improved to realize the extraction of human eye feature points and rough positioning of pupils.Then,aiming at the pupil region image obtained after rough positioning,based on the statistical learning method and the pupil edge extraction method,combined with the deformable template method,a pupil precise location algorithm based on sub-pixel contour shape multi-template matching is developed.The experimental results show that the proposed method can accurately and quickly locate the center coordinates of the human eye pupil,and has good robustness against phenomena such as eyelid occlusion and changes in external light.(3)Taking the horizontal diameter of iris as the ranging scale,a method for measuring the horizontal diameter of iris based on gray integral projection and gray gradient change is proposed.Using the principle of keyhole imaging and pinhole camera model,a single visual distance model was established,and the distance between the pupil of the human eye and the camera was calculated by the horizontal diameter of the iris to achieve the purpose of pupil distance measurement.(4)Apply the above method to infrared facial images collected by the image acquisition platform.Firstly,the camera internal parameters were obtained through calibration experiments.Subsequently,infrared facial images were collected and the pupil plane localization algorithm proposed in this paper was used to locate the human eye pupil.The experimental results showed that the algorithm had high accuracy and could meet the accuracy requirements of ophthalmic disease detection instruments for pupil localization.Finally,the pupil ranging algorithm proposed in this article can improve the measurement range of ophthalmic disease detection instruments.The comprehensive experimental results indicate that the method proposed in this article has high accuracy and stability,and can be used in the field of ophthalmic medical equipment.