Research Of Associative Camera Surveillance System Based On Face Landmarks Detection

With the rapid development in computer vision and hardware technology, face recognition has been widely used in many areas, especially in the application of video surveillance. While in many existing video surveillance systems, the camera cannot get the high resolution image due to hardware, cost and environment constraint, which is the reason that the face recognition technology cannot be applied well. A new associative web camera system is presented in this paper to solve this problem, which includes face landmarks detection algorithm, associative web camera system design and relevant application realization.Firstly, unlike the usual approach, we applied the object detector based on Deformable Part Model to detect face landmarks. This method performs well in execution efficiency and accuracy, obtaining more than90%accurate with less than ten milliseconds’ time cost. Based on intensive research, we give performance comparison results of LBP pyramid, LBP uniform pattern and LBP directed gradient feature. In these three features, LBP uniform pattern achieves a high accuracy with the minimum time cost, which is most suitable for the associative camera surveillance system. Besides, a face detection post-processing algorithm is proposed using the result of face landmarks detection, which improves the face detection accuracy a lot.Secondly, a geometric space model of reference camera, steered camera and the object individual is proposed to obtain steered camera’s magnification and steering parameters. The model also utilizes the distance of face landmarks and the statistics of that distance. In order to achieve the best performance, the system design was optimized in many aspects after realization and testing.Thirdly, two applications based on face recognition technology are realized to meet the present requirements in this field. One is interactive picture and video searching system based on face recognition, and another one is the face recognition surveillance system used in public areas like the exhibition hall, which can give the track of the specified person within the field of camera vision in real-time.