Real-Time Human Detection and Counting System Using Human Model and Axial Direction Filtering
友宏 中上, Hiroshi Matsuo · IEEJ Transactions on Electronics Information and Systems · 2002
We propose new fast human detection method and counting system corresponding to the environment where many persons exist in camera view. We develop a human model from the feature of region change of many images of pedestrians which walk toward various directions, and detect two or more persons who are overlapping by matching with difference image and human model. And we propose the axial direction filtering for simplification of matching process. In counting system, we also propose the algorithm which counts measurement line passage of pedestrian using information of passage position and body color. The experiment in the environment where many persons exist in camera view shows that the proposed systems are possible to work in real-time with sufficient accuracy.