Real-Time Human Motion Detection and Tracking with Learning based Representation

Abstract

Nowadays, real-time information is very important and learning based human motion has fascinated range from detection to tracking state in Computer Vision. In this system, the real-time videos are used to detect, track, and classify object or events in order to understand a real-world scene. Video based real time human motion detection and tracking is a complex and challenging task due to variation in human pose, shape variation, illumination changes and background appearance. A real-time mechanism is to detect the person and their moving within an environment from the video camera. This paper proposes human motion detection from video sequences. The proposed method includes three stages: human detection, motion tracking and accuracy result based on learning approach. The result is to become an efficient detection system for real-time human motion. Motion detection and tracking is determined by using Histogram of Oriented Gradients (HOG) feature extractor and Support Vector Machine (SVM) detector with learning human pattern which is well performed human detection and tracking in video sequences. Detailed analysis is carried out on the performance and accuracy of the system with the various test videos to show the results. The experimental results demonstrate the efficiency of the method