Pedestrian recognition using automotive radar sensors

[摘要] The application of modern series production automotive radar sensors topedestrian recognition is an important topic in research on future driverassistance systems. The aim of this paper is to understand the potential andlimits of such sensors in pedestrian recognition. This knowledge could beused to develop next generation radar sensors with improved pedestrianrecognition capabilities. A new raw radar data signal processing algorithm isproposed that allows deep insights into the object classification process.The impact of raw radar data properties can be directly observed in everylayer of the classification system by avoiding machine learning and tracking.This gives information on the limiting factors of raw radar data in terms ofclassification decision making. To accomplish the very challengingdistinction between pedestrians and static objects, five significant andstable object features from the spatial distribution and Doppler informationare found. Experimental results with data from a 77 GHz automotive radarsensor show that over 95% of pedestrians can be classified correctly underoptimal conditions, which is compareable to modern machine learning systems.The impact of the pedestrian's direction of movement, occlusion, antenna beamelevation angle, linear vehicle movement, and other factors are investigatedand discussed. The results show that under real life conditions, radar onlybased pedestrian recognition is limited due to insufficient Doppler frequencyand spatial resolution as well as antenna side lobe effects.

[学科分类] 电子、光学、磁材料