Research On Face Detection And Tracking Technology Based On OpenCV

Face recognition is widely used in intelligent monitoring and access control systems,which promotes the rapid development of computer vision and intelligent information processing.Face detection and tracking technology plays an indispensable role in the development.However,the complexity of real scene makes the positioning of human face vulnerable to illumination variation and noise.At the same time,the uncertainty of the face posture also increases the difficulty of face detection.To solve this problem,face detection and face tracking methods are respectively studied in this thesis,and then the two methods are exploited to implement the accurate positioning of human faces.The following are the main research contents:(1)Face detection method.First of all,image preprocessing is performed on the images to reduce the interference of feature extraction such as illumination variation and noise,so as to improve the effect of face detection.Secondly,the face detection algorithm that extracts haar-like features and cascades AdaBoost classifiers is used to detect human faces.But the traditional AdaBoost algorithm can only detect front faces and fail to detect side faces.In order to solve this problem,the frontal face detector is used to detect firstly,and the side face detector is used to detect the images secondly,then the multi-angle face detection algorithm based on AdaBoost is finally constructed.According to the experimental analysis,the face detection algorithm has higher detection accuracy than the traditional AdaBoost algorithm.(2)Face tracking method.By using the circulant matrices and kernelized correlation filtering ideas,the Kernelized Correlation Filter algorithm increases the tracking speed to hundreds frames per second and achieves better tracking results.However,when the face are affected by such as fast motion,motion blur,occlusion,and similarity,the KCF algorithm does not consider the motion state information of the face,resulting in the failure of tracking.In order to solve this problem,a new face tracking algorithm named CNK-KCF is proposed.In the tracking process,the Kalman filter is used to predict the coordinates of face position in the next frame,and the tracking window is selected from the predicted coordinates.Then use the KCF that extracts CN features to detect the face coordinates,and finally use the detection results to update the Kalman filter.According to the experimental results,the CNKKCF tracking algorithm has better robustness than other advanced tracking algorithms.(3)By combining the face detection method and the face tracking method,this thesis constitutes an overall system framework for face detection and tracking.Firstly,OpenCV is configured on VS2010 as the experimental platform.Secondly,apply this image preprocessing method to preprocess the image and utilize the multi-angle face detector to detect the image.Then initialize the tracker by using the face position of the current frame.Finally exploit the CNK-KCF tracking algorithm to continue to locate human face.By this way,the face detection and tracking technology based on OpenCV is constructed on this experimental platform,which can achieve accurate detection of multiple faces and multiple angles,and robust tracking of single faces.