Radiometric Temperature-Based Pedestrian Detection For 24 Hour Surveillance

摘要

This demo presents a novel pedestrian detection method using radiometric temperature information for 24 hour operational capability. Conventional surveillance systems usually use brightness information of thermal images, which leads to unstable pedestrian detection due to the unwanted signature distortion by hot background spot. In this demo, calibrated temperature is used instead of auto contrasted brightness. Pedestrian temperature shows reduced visual variations compared to the brightness value. The physical temperature is normalized by 40 degrees Celsius that is the highest human temperature, which can further increase the detection performance. Current demo system uses the ACFboosted tree detection platform to validate the effectiveness of radiometric temperature. Several demo videos demonstrate higher detection rate and lower false alarm rate.