A New Method for People-Counting Based on Support Vector Machine

This paper proposed a new method for infrared people-counting. According to the characteristics of the time continuous data collected by infrared sensors, the pattern recognition idea is introduced in the data processing procedure. After the special pretreatment, adaptive segmentation and feature extraction for the people-counting data, the feature vector is used as the inputs of the trained support vector machine classifier to classify and statistic the total number of the people who go through the infrared sensor area in a period of time. Compared with traditional people-counting using sensor, this method is more accurate and it can count the people number at the situation that several people go through the infrared sensor at the same time. Finally, the experiments indicate that the method can be applied in actual application.