A Method for Counting Passing People Using Person Detection Based on a Combination of Difference Images and Region Growing

Abstract

In this study, we propose a method to measure the number of people based on a bird’s-eye movie. Local Binary Patterns (LBP) features are calculated for the input video, and the person area is estimated based on the changes of LBP features between frames. Using the estimated person area as a starting region, a region expansion method using background difference and RGB values is applied to obtain the person area. If the person area is composed of multiple people, each individual is detected using the past detection results and the K-means method. The moving direction of the passing people is determined based on the passing order on two virtual lines. As the result of the experiments, we have confirmed the effectiveness of the people detection and the measurement methods.