Date of Publication :7th May 2016

Abstract: Human tracking is an important research area in computer vision with potential applications in many other fields like augment reality, human–machine interaction, and advanced driver assistance systems. Despite a lot of progress in the field, visual tracking remains a difficult problem due to many challenges. The proposed work provides a solution for multiple human tracking methods, which uses a combination of Blob extraction and Kalman filtering. The main objectives of the proposed work are to precisely track moving or static human beings and to identify and estimate their future location in an unknown scene. A pre recorded input video is split into individual frames which are actually processed to track the human beings and recognize their actions. The background modeling is done using Blob extraction followed by Kalman filtering & Optical flow to detect the positions of humans & track them. HOG features are used to classify and recognize various human actions. The results are the frames of the video file, which consists of marking the borders of the human beings appearing in the video, tracking them and displaying their actions such as walking, sitting and throwing on the command window of Matlab.

Reference :