A Methodology Review on Multi-view Pedestrian Detection

Although surprisingly good progress has been made in monocular pedestrian detectionPedestrian detection with emerging deep learning techniques, the existing algorithms still suffer from heavy occlusion between people. To cope with this problem, multiple cameras are often used to provide complimentary information. This chapter presents a comprehensive review on the algorithms for multi-view pedestrian detectionPedestrian detection. The existing methods are sorted into low-level fusion, intermediate-level fusion, high-level fusion and deep-learning based fusion, in terms of the degree of information fusion. The challenges in developing deep multi-view algorithms are also described in this chapter.