Experimental Research on NLOS Area Monitoring  

Background

For vehicles and mobile robots, detecting pedestrians and obstacles in non-line-of-sight areas is an important problem. Accident risk increases when pedestrians suddenly appear from behind parked vehicles, building corners, walls, or intersections.

Cameras and LiDAR have difficulty observing objects when the line of sight is blocked. In contrast, radio waves used in radar systems may include information about non-line-of-sight targets through diffraction, reflection, scattering, and multipath propagation, depending on the environment. Although this does not mean complete see-through sensing, radar may detect the presence and motion tendency of humans in blind areas.

In this research topic, the goal is not only to detect humans but also to classify motion patterns related to safety, such as approaching, passing, or rushing out from blind areas.

Research Objectives

This research aims to develop basic radar sensing techniques for detecting humans and pedestrians in blind areas and classifying their motions.

Main topics include:

• Detection of pedestrians in non-line-of-sight areas occluded by walls or vehicles

• Motion detection of humans approaching from blind areas

• Classification of motion patterns related to rush-out risk

• Human-motion discrimination using micro-Doppler features

• Experiments in indoor and outdoor blind-area environments

• Analysis of radar observations including diffraction, reflection, and multipath components

• Applications to vehicle safety systems and mobile robots

Related Work

Our previous work has investigated pedestrian detection and motion classification in blind areas formed by vehicles in outdoor environments and by walls in indoor environments. Doppler information and micro-Doppler features obtained by radar are used to detect pedestrians in blind areas and classify motion patterns related to rush-out risk.

This series of studies can be regarded as early experimental work on radar-based human monitoring in blind areas. In particular, it experimentally examined pedestrian motion and rush-out risk in blind areas using micro-Doppler radar under both indoor and outdoor conditions close to realistic situations.

Even when a region cannot be directly observed by optical sensors, information contained in diffracted waves, reflected waves, and multipath components may be useful for human detection and safety support in blind areas. Future work includes experiments under more realistic conditions, use of range-angle information from FMCW or MIMO radar, and sensor fusion with other sensing modalities.

Selected Publications

• Sora Hayashi, Kenshi Saho, Daiki Isobe, and Masao Masugi, “Pedestrian Detection in Blind Area and Motion Classification Based on Rush-Out Risk Using Micro-Doppler Radar,” Sensors, 2021. 

• Sora Hayashi, Kenshi Saho, Hiroaki Okinaka, Lin Meng, and Masao Masugi, “Detection and Classification of Human Motion in Blind Area Using Micro-Doppler Radar: Fundamental Experiments Toward the Prediction of Dash-out from Blind Area,” ICAMechS, 2019.
  Summary: This study conducted fundamental experiments on human motion detection and classification in blind areas using 24-GHz micro-Doppler radar. It is positioned as an early study leading to the later Sensors paper. 

• Ryosuke Iida, Kazuhide Kamiya, Xiangbo Kong, and Kenshi Saho, “Classification of Rush-Out Risk of Pedestrians in Blind Area Using 2.4 GHz FMCW Radar,” ATAIT, 2023.

Keywords

blind-area monitoring, blind-area pedestrian detection, rush-out risk classification, rush-out prediction, non-line-of-sight area, NLOS sensing, micro-Doppler radar, millimeter-wave radar, FMCW radar, radio-wave diffraction, reflection, multipath propagation, pedestrian detection, vehicle safety, mobile robot, radar signal processing