Pedestrian detection with image segmentation and virtual mask

Pedestrian detection is one of the most popular research areas in video processing and it is vital for video surveillance systems. In this paper, we present a real-time pedestrian detection system based on Dalal and Triggs´s human detection framework with the use of image segmentation and virtual mask. Image segmentation enables the system to focus only on the region of interest whereas the virtual mask reduces noise and unnecessary processing. The system has high computational speed while it retains the discriminative power of Histogram of Oriented Gradient (HOG) features for pedestrian detection. It is tested with videos captured in a single task environment and in various lengths. The experimental results showed that most of the pedestrians in the videos are correctly detected, achieving an overall accuracy of 83.29% with low amount of computational time, which implies the effectiveness in detecting pedestrians and the real-time property of the proposed system.