Design And Research On Hand Held Pupil Detector

Pupil detection is one of the commonly used methods in clinical diagnosis,which provides effective basis for patient-health such as drug taking,intracranial illness,and mental state and so on.With the rapid development of embedded technology and information technology,embedded handheld devices are widely used in medical work in recent years.Hand held pupil detector makes the machine has ability of environment understanding by using image processing technology,which regards computer vision technology as the core.And it has the characteristics of compact,fast and convenience of man-machine interaction when measuring and locating pupil with embedded device,which has broad application prospects.In this paper,a kind of embedded Linux hand held pupil detector based on ARM9 architecture is mainly studied,under the non-contact condition,which is used in pupil recognition and localization through fixing the distance between the detector and the measured object by equipment.In hardware system,S3C2440 A is used as the core,OV9650 camera is selected to acquire image,and the LCD is adopted to operate software and output data.In software system,embedded Linux is selected,system and software transplantation is completed by cross compiling,and interactive desktop is created by QT/Embedded to achieve the display and measurement of pupil data.The pupil is measured by Hough and mass labeling based on openCV by development of pupil detection algorithm respectively.Circle detection is widely used in image processing.The pupil detection is a special application in circle detection.Two kinds of pupil measurement algorithms of Hough detection and mass labeling are introduced in this paper in detail based on embedded system design.Hough pupil detection: Firstly,graying the color image.Then canny edge detector is used to extract the edge of the eye.In edge extraction process: First,smoothing the image noise by Gauss filter.Second,canny operator is used to enhance the eye edge points.Third,non-maxima is used to extract the image edge points.Fourth,the double threshold method is used to detect the edge of the edge points.Finally,the Hough transform is used to identify and detect the pupil contour to locate pupil and display the results on the screen.Mass labeled pupil detection: Firstly,the gray image is transformed into two values by using the optimal threshold segmentation.Then,the two value image after inversion is marked by mass labeling to select the mass that meet the requirements.Finally,according to the feature between area and perimeter of circle,the mass labeling method proposed by F.Chang et al is selected to trace mass profile and recognize pupil.The results are displayed on the screen and the pupil radius and time curve is drew to observe the changes of pupil.Experiments show that Hough pupil detection algorithm has high precision,low noise,but higher system delay,which can achieve the desired effect and especially suitable for the precise measurement of pupil at a certain moment.And the detection accuracy of the mass labeling method is lower than that of Hough,but the real-time detection is better,which has a good effect on the continuous monitoring of the pupil for a period of time.