Carrying object detection and tracking based on body main axis

In this paper, we propose a method to examine whether a walking person is carrying objects using the width of human body contour. First we concentrate on the detection of the moving people contour and then human contour width is projected to the feature space for training a classifier. We introduce Principal Component Analysis for dimension reduction and Support Vector Machine to classify the carrying object status. If the status is true, the carrying object region is figured out by analyzing human contour shape. We utilize the outmost vertex from human contour to track the carrying object such as a rucksack or luggage. The experimental results demonstrate that our approach is encouraging.