Real-time Eye Detection Method Robust to Facial Pose Variations Using Gradient Directional Features and Particle Filter

This paper describes the iCare Interaction Assistant, an assistive device for helping the individuals who are visually impaired during social interactions. The research presented here addresses the problems encountered in implementing real-time face recognition algorithms on a wearable device. Face recognition is the initial step towards building a comprehensive social interaction assistant that will identify and interpret facial expressions, emotions and gestures. Experiments conducted for selecting a face recognition algorithm that works despite changes in facial pose and illumination angle are reported. Performance details of the face recognition algorithms tested on the device are presented along with the overall performance of the system. The specifics of the hardware components used in the wearable device are mentioned and the block diagram of the wearable system is explained in detail.

The use of video sequences for face recognition has been relatively less studied than image-based approaches. In this paper, we present a framework for face recognition from video sequences that is robust to large changes in facial pose and lighting conditions. Our method is based on a recently obtained theoretical result that can integrate the effects of motion, lighting and shape in generating an image using a perspective camera. This result can be used to estimate the pose and illumination conditions for each frame of the probe sequence. Then, using a 3D face model, we synthesize images corresponding to the pose and illumination conditions estimated in the probe sequences. Similarity between the synthesized images and the probe video is computed by integrating over the entire sequence. The method can handle situations where the pose and lighting conditions in the training and testing data are completely disjoint.

Eye detection is the most important and critical task of diverse applications such as face detection and recognition. However, most eye detection methods do not fully consider detection robustness to people with glasses, illumination variation, head pose change, and eye occlusions. This paper proposes an efficient hybrid eye detection method based on a gray intensity variance filter (VF) and support vector machines (SVMs). Firstly, the VF is used for eliminating most of noneye region images to keep less candidate eye regions. Then accurate two eye regions are determined easily through the trained SVM classifier. Moreover, this paper provides an assessment of the sensitivity of obtained parameters in the SVM classifier on eye detection accuracy. The proposed method was evaluated on different face databases. The experimental results show that the method can improve the performance of eye detection and achieve a higher detection accuracy compared with state-of-the-art methods.

Face and eye detection is one of the most challenging problems in computer vision area. The goal of this paper is to present a study over the existing literature on face and eye detection and gaze estimation. With the uptrend of systems based on face and eye detection in many different areas of life in recent years, this subject has gained much more attention by academic and industrial area. Many different studies have been performed about face and eye detection. Besides having many challenging problems like, having different lighting conditions, having glasses, facial hair or mustache on face, different orientation pose or occlusion of face, face and eye detection methods performed great progress. In this paper we first categorize face detection models and examine the basic algorithms for face detection. Then we present methods for eye detection and gaze estimation.

Focused on the issue that the detection error rate of the current eye detection method is relatively high, and when the AdaBoost algorithm is used to train the classifier, it is easy to appear the phenomenon of weight imbalance. A new eye detection method based on the improved AdaBoost algorithm is proposed. First, the AdaBoost algorithm is applied to the detection of human eyes. Then the reason for the imbalance of weights in training of AdaBoost algorithm is analyzed, and the concept of missing detection rate is introduced to improve the weight updating process of AdaBoost algorithm. The experimental results show that the improved AdaBoost algorithm ensured the sample weight distribution balance and improve the accuracy in the training process; eye detection based on the improved AdaBoost algorithm effectively maintains a high detection rate and inhibits the detection error rate, makes detection more accurate.

An innovative method for cascade eye detection using facial geometric features is proposed. The cascade detector is used to detect pairs of eye of candidates. The size similarity, the horizontal symmetry, the horizontal sextant angle, and the scale coefficients extracted from facial geometric feature are integrated to improve eye detection. Both the BioID database and real-time video stream were used in the experiments. Experimental results show that the proposed eye detection method is significantly more accurate than the cascade detector alone without sacrificing overmuch time-efficiency.

In order to investigate whether or not checkpoints of features surrounding the eye are suitable for eye tracking, we evaluated the checkpoints using statistical hypothesis testing. By combining the checkpoints of features surrounding the eye, the eye tracking system using the template matching was effective for the eye detection. The parameters of the checkpoints were empirically determined to detect the eye in any individual. The statistical hypothesis testing revealed that the empirical setting values of the checkpoints were significantly different from actual facial features obtained using face images. We investigated suitable parameters of the checkpoints derived from actual facial features obtained with facial images. It was revealed that new checkpoint settings were effective in drastically reducing the possibility of incorrect eye detection, although the correct detection rate was slightly decreased.

This paper presents the evaluation of visual features for the proposed two eye detection method applied to thermal images. The use of two eye region is due to its distinctive pattern and to overcome the issue of blurred and noisy characteristic in the thermal image. Comparative performance analysis on three different features which includes Haar, Histogram of Oriented Gradients (HoG) and Local Binary Patterns (LBP) is conducted. The performance of the eyes detection method is measured based on the correct detection of both eyes inside the face image. The experiments were done on the Natural Visible and Infrared Facial Expression Database (NVIE). The method proposed in this paper shows good eye detection accuracy. The best detection accuracy is obtained using the HoG feature.

In refractive eye surgery using laser, automatically localizing a target eye for surgery and precisely focusing the laser to the corneal surface are crucial to the medical treatment. We therefore propose a new eye detection method based on template matching for automatic localizations of target eyes for refractive eye surgery with excimer laser. In the proposed approach, representative templates of image gradient histograms of eyes are constructed using both open and privately-collected eye datasets. Real-time eye detection that helps guiding the excimer laser to a target eye for surgery can be achieved by the efficient template matching. After the eye localization, image-guided, automatic alignment of positioning marker, i.e., a read line of visible light, is then employed to measure the distance between the corneal surface and the excimer laser. Experimental results show that the proposed eye detection method using template matching outperforms an HoG-based eye detector in miss detection rates based on the same training data for this specific medical application. Also, with the automatic image-guided positioning of laser head, the time consuming process of manual laser focusing can be largely simplified.

A set of transforms (SVD transforms) were introduced in (Shahshahani and Tavakoli Targhi) for understanding images. These transforms have been applied to some problems in computer vision including segmentation, detection of objects in a texture environment, classification of textures, detection of cracks or other imperfections, etc. This technique is shown to be applicable to determination of the location of eyes in a facial image. This method makes no use of color cues, prior geometric knowledge or other assumptions and does not require training. It is also insensitive to local perturbations in lighting, change of orientation and pose, scaling, and complexity of the background including indoor and outdoor environments. The method can be used for eye tracking and has applications to face recognition. It has also been used in animal eye detection and differentiation. © 2007 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 16, 222–229, 2006

In this paper, we propose a practical method for eye blink detection using a monocular system. Eye blink detection is an important technology in many situations such as facial action analysis and signal processing. However, automatic eye blink detection is quite difficult since the eye blink is occurred fast. Fast head tracking and eye detection methods are therefore required in real environment. The proposed method is based on the facial feature point tracking algorithm by using particle filtering. The proposed method works at over 60fps and is able to detect accurate blink motion in real time although our system require only one USB camera. In order to confirm the effectiveness, we show some of applications using our blink detection system.

The process of eye tracking has been subject of intensive research during the last years. The main aim consists in the determination of the point of gaze or the eye movement relative to the head. In this work, we use a static video camera to realize the tracking of the eyes movement mainly due to its important features such as low cost, fully automated and easy application. The proposed approach encompasses a solution to the problems of eye detection and eye tracking. Our eye detection method employs a set of cascaded Haar classifiers that enable the fast recognition of eye regions in a real-time sequence of images. On the other hand, the eye tracking method uses particle filtering that is robust in clutter environments and allows the fusion of sensor data as well as the tracking of nonlinear systems. The results demonstrate that the eye tracking system is capable of robustly track the eyes under variable light conditions and under several different face orientations.

Quadrotor is one type of unmanned aerial vehicle that has the ability to vertical takeoff and landing. In this research, a system designed to stabilize quadrotor during flight condition by maintaining at angle of roll, pitch, yaw, and x, y, and z axis position using LQR full state feedback with artificial neural network (ANN).The LQR full state feedback method uses 12 states with each K constant being tuned with ANN. This research implements ANN method to change feedback constant at angle of roll, pitch, and yaw and x, y, and z axis. The artificial neural network method uses 12 input layers, 12 hidden layers, and 1 output layer.Testing with ANN improved the rise time to ± 2.18 seconds at the roll angle, ± 1.23 seconds at the pitch angle, and ± 0.31 seconds at the yaw angle. Improved settling time value up to ± 2.41 seconds at roll angle, ± 1.23 seconds at pitch angle, and ± 1.07 seconds at yaw angle. Improved steady state eror value of ± 0.61% at roll angle, ± 4.88% at pitch angle, and ± 0.82% at the yaw angle.

Eye detection can be used in intelligent human-computer interfaces, driver drowsiness detection, security, and biology systems. In this paper a new method for eye detection based on rectangle features and pixel-pattern-based texture feature (PPBTF) is proposed. First, Adaboost cascade classifier by rectangle features is constructed to do rough eye detection in a front facial image. Second, the result image patches are cropped and scaled to 24×12 to compute the features of PPBTF, then, put these features into an Adaboost and SVM classifier for an accurate detection. Some eye open front facial images from FERET Database are chosen to do the experiments, and the results show that the approach is real time and is effective for eye detection.

We developed a signal processing algorithm to determine three components of the velocity vector of a highly maneuverable aircraft. We developed an equation of the distance from an aircraft to an underlying surface. This equation describes a general case of random spatial aircraft positions. Particularly, this equation considers distance changes according to an aircraft flight velocity variation. We also determined the relationship between radial velocity measured within the radiation pattern beam, the signal frequency Doppler shift, and the law of the range changing within the irradiated surface area. The models of the emitted and received signals were substantiated. The proposed equation of the received signal assumes that a reflection occurs not from a point object, but from a spatial area of an underlying surface. It fully corresponds to the real interaction process between an electromagnetic field and surface. The considered solution allowed us to synthesize the optimal algorithm to estimate the current range and three components {Vx,Vy,Vz} of the aircraft’s velocity vector V→. In accordance with the synthesized algorithm, we propose a radar structural diagram. The developed radar structural diagram consists of three channels for transmitting and receiving signals. This number of channels is necessary to estimate the full set of the velocity and altitude vector components. We studied several aircraft flight trajectories via simulations. We analyzed straight-line uniform flights; flights with changes in yaw, roll, and attack angles; vertical rises; and landings on a glide path and lining up with the correct yaw, pitch, and roll angles. The simulation results confirmed the correctness of the obtained solution.