15-04-2017, 04:58 PM
A technique for eye detection and gaze estimation was developed in MATLAB for omnipresent cases in which a user sits in front of a camera and a screen. Such a technique could enable computer interfaces that determine the user input of low-effort eye movements that naturally express the user's intention. In addition, applications such as evaluation of driver fatigue, surveillance or advertising would benefit from this technology. The location of the eyes in the images is a problem previously solved; However, "black box" eye detection techniques do not allow us flexibility and understanding of the basic functionality required for a look-oriented interface. To evaluate the effectiveness of this technique two experiments were performed. First, the ocular detection (location of the iris centroid and the pupil) was performed in images of the training set with 50% success in finding both eyes with zero false positives, 94% success in finding at least one eye Subject and 17% Success in finding both eyes of subjects who do not wear glasses.
Human eye tracking and look estimation is a rapidly growing area of research with many useful human interaction applications such as video conferencing and eye writing. Eye tracking can also be used in monitoring the driver's alertness by observing the appearance of the eyes for drowsiness or fatigue. Improving eye tracking and eye tracking systems also provides a means of substitution for physically handicapped people who find it difficult or are completely unable to control a mouse. In this research, we focused on the development of a cheap and non-intrusive look estimation system. The proposed system requires little calibration for new or return users and is robust against normal lighting conditions and moderate head movements. Non-intrusive research in ocular detection, tracking, and gaze estimation can be classified into three categories: passive-based approaches, active infrared-based approaches, and hybrid approaches. Passive-based approaches detect and track the eyes by exploiting the unique intensity distribution (eg, the dark pupil and the white sclera) or the shape (eg the circular iris and the angles of the eyes) of the eyes to Distinguish the human eye from other objects. Three typical methods are template-based, appearance-based, and feature-based. These traditional methods based on passive images obtain decent ocular follow-up results for good contrast images when the faces are in the frontal orientations and the eyes are without closure and occlusion. However, they may not work well for different topics under different lighting. Active infrared based approaches exploit the spectral (reflective) properties of the pupils under close infrared illumination to produce the bright / dark pupil effect for the detection and tracking of the eyes. However, these methods require a clear / dark clear pupil effect to function well and rely heavily on the brightness and size of the pupils, which are often affected by ocular closure and occlusion due to the rotation of the face , External illumination interferences and the distances of subjects to the camera. In addition, they require a sophisticated control system and an expensive camera capable of generating interlaced images using even and odd fields.
Human eye tracking and look estimation is a rapidly growing area of research with many useful human interaction applications such as video conferencing and eye writing. Eye tracking can also be used in monitoring the driver's alertness by observing the appearance of the eyes for drowsiness or fatigue. Improving eye tracking and eye tracking systems also provides a means of substitution for physically handicapped people who find it difficult or are completely unable to control a mouse. In this research, we focused on the development of a cheap and non-intrusive look estimation system. The proposed system requires little calibration for new or return users and is robust against normal lighting conditions and moderate head movements. Non-intrusive research in ocular detection, tracking, and gaze estimation can be classified into three categories: passive-based approaches, active infrared-based approaches, and hybrid approaches. Passive-based approaches detect and track the eyes by exploiting the unique intensity distribution (eg, the dark pupil and the white sclera) or the shape (eg the circular iris and the angles of the eyes) of the eyes to Distinguish the human eye from other objects. Three typical methods are template-based, appearance-based, and feature-based. These traditional methods based on passive images obtain decent ocular follow-up results for good contrast images when the faces are in the frontal orientations and the eyes are without closure and occlusion. However, they may not work well for different topics under different lighting. Active infrared based approaches exploit the spectral (reflective) properties of the pupils under close infrared illumination to produce the bright / dark pupil effect for the detection and tracking of the eyes. However, these methods require a clear / dark clear pupil effect to function well and rely heavily on the brightness and size of the pupils, which are often affected by ocular closure and occlusion due to the rotation of the face , External illumination interferences and the distances of subjects to the camera. In addition, they require a sophisticated control system and an expensive camera capable of generating interlaced images using even and odd fields.