Study On The Technology Of Detection And Tracking Of The Face

Moving target detection and tracking has been an important point of intelligent video monitoring work for a long time.Detecting and tracking moving human face is a strong theoretical and applied research branch in the scene of lighting.But there are still much of problems need to be solved in this area.Experimental data show that only from the bottom of the image to do analysis could not accurately extract the target of segmentation.Especially when the illumination problem is the prominent confusion, shadows and highlights greatly weaken the results’ accuracy.Therefore,combined with the experimental data, in the image sequence, for illumination problems under the scenario of face detection and tracking is studied. The Intrinsic Face Detection and tracking(IFDT) algorithm is proposed in this paper, which has two main parts.The first part is mainly to reduce the scope of face detection and tracking. It finds containing face and has filtered light candidate regional influence from a complete image.The sequence image is divided into two times to get rid of the light and well testing tracking effect:Combined gaussian background modeling with the light has nothing to do with graph theory, Using the double difference fast and efficient removal of follow the movement of the human body silhouette.On the basis of the theory of the overall segmentation of human movement, and getting the body area of the segmentation,and finding the candidate area of the face detection, which greatly accelerated aging test track.Removing the influence of light against the candidate area of secondary.It is the basic work necessary for positioning the next step.The second part is the specific process and content determine face IFDT methods in the location and regional describes.The ingenious combination of binocular gradient feature and the ellipse template, and multiple screening mechanism and testing the treated area.Location using binocular characteristics and determine the facial region with the ellipse template, with signature verification finally,those ensures the accuracy of the detection of trace.