| With the continuous improvement of socioeconomic level and the development oftechnology of computer vision, intelligent video monitoring system application prospect isalso more and more widely, a large number of monitoring system has been used in thestation, roads, supermarket entrances, etc... Intelligent video surveillance system is thefuture direction of the video surveillance system. Through pedestrian detection andtracking in the video monitoring system to count the number has very important practicalsignificance. The number of statistical methods and techniques in the video system willbecome difficult and hot in computer vision research.According to the process of people counting on video monitoring, analyzed themoving target segmentation, detection and tracking algorithm both in China and foreigncountries. Taking into account the case in close video surveillance, the human face is themost obvious target area, so face detection and tracking are used to count the number ofpedestrian.Face detection, proposed a detection method based on skin color and facial feature.First, image preprocessing, such as light reinforcement, image smoothing operation, thentransformed the image from the RGB mode into the YIQ mode, separated skin regionsthrough the threshold. Through Trained classifier on Skin regions to proceed facedetection. This method accelerated the detection rate and reduced the false detection rate.Face tracking, proposed a multiple target tracking method based on multi-featurefusion, that is, through the image texture feature space and color feature space adaptiveweighted fusion for target tracking. First, calculated the probability distribution diagramsrespectively of texture features and chroma features, and then weighted to get a newprobability distribution, then using Camshift algorithm for face tracking.Goal to improve real-time performance and veracity of people counting, a goodmethod was proposed, which achieve good results in the experiment. |
PDF Full Text Request