Human Eye Interest Point Detection Based On Eye Tracking

Eye tracking technology has been a hot spot in the fields of human-computer interaction,psychological testing,virtual reality and artificial intelligence.The eyes are the form of expression of human psychology and thoughts in appearance,and eye tracking technology can reflect human psychological changes by extracting human eye's information.The key point of this paper is to extract human eye information by using computer vision and image processing technology.Based on the traditional eye tracking,the main innovations in this paper are as follows:.1.In this thesis,an eye tracking method based on deep learning is proposed.The network adopted in this method is a new one with three convolutional layers,three pooling layers and two fully connected networks.The convolutional neural network(CNN)can be used to detect the location of key points of human eyes succinctly and efficiently,with an accuracy rate of 0.88.Experimental results show that this model is more accurate than traditional cascade classification.2.In this thesis,a new method of extracting pupil and infrared light spot was proposed to improve the accuracy of traditional threshold processing.This paper combines edge detection with threshold processing.Based on prior knowledge of pupil position,we can ignore the gradient value of image edge.The maximum point of the coordinate of the gradient point is determined by the gradient operator,and the area of the interested rectangle with the maximum value as the center is determined.Threshold processing in the interesting of region can better avoid interfaceing of environmental and noise.3.In this thesis,the interest point detection algorithm based on eye tracking is applied to the interest test system to estimate the fixation points of human eyes in real time.The system judges the people's interests and hobbies by detecting the changes of eye fixation points and facial expressions at the same time,which demonstrate the algorithm has a good application value.4.In this thesis,based on the array camera,a super resolution algorithm is proposed,this algorithm can better improve the resolution of the image.The same time,multiple-lens can realize the three-dimensional space coordinates detection and overcome shortcomings that human should remains head unchanged and the distance of head and lens is short under two-dimensional,the array camera will apply to the eye tracking in the future.