A Real-time Pedestrian Detection System Based On Video Surveillance

Moving object detection based on images and video is one of the most fundamental questions in the field of computer vision and artificial intelligence.Pedestrian detection plays a key role for a modern intelligent transportation system and video surveillance system.However,due to the characteristics of pedestrians,such as non-rigid and diversity,accuracy and speed are still research challenges.In this paper,we present a real-time video-based pedestrian detection system,which can satisfy the demands of real-time and accuracy in video surveillance system.In this paper,we propose a robust and fast pedestrian detection system,which combines an improved people counting scheme and an effective ROI extracting method.To improve the degree of accuracy,we use a dimensionality reduction method which combines multi-scale local binary patterns(MLBP)features and normalized histograms of oriented gradient(HOG)features with Principle Component Analysis(PCA).To accelerate the detection speed,we propose a reliable ROI extracting method to reduce the detection area.ROI extraction reduces the number of detection windows,resulting in a signification reduction in detection time.Extensive experiments and analyses show that our method outperforms state-of-art techniques for both speed and reliability.Our proposed pedestrian detection system can correctly detect the positions of pedestrians in real time.