Human Detection Using Thermopiles

Abstract

This paper presents a simple approach to detect people using an 8times1 thermopile array sensor. The performance of the sensor is evaluated in terms of detection rates and false positives obtained in the application of the sensor to people counting at doorways and pedestrian detection from a mobile robot in a real environment. To ensure that people is detectable regardless of the ambient temperature, the proposed method normalizes the temperatures according to a model for the maximum expected skin temperature as a function of the ambient temperature. The working principles and characterization of the thermopile sensor are also presented. This should be of interest and pedagogical value to many roboticists who wish to equip their robot with a rather effective passive infrared sensor, which unlike vision-based methods is not affected by changes in the illumination conditions of the environment.