The presence of Non-Line-of-Sight (NLoS) blind spots resulting from roadside parking in urban environments poses a significant challenge to road safety, particularly due to the sudden emergence of pedestrians. mmWave technology leverages diffraction and reflection to observe NLoS regions, and recent studies have demonstrated its potential for detecting obscured objects. However, existing approaches predominantly rely on predefined spatial information or assume simple wall reflections, thereby limiting their generalizability and practical applicability. A particular challenge arises in scenarios where pedestrians suddenly appear from between parked vehicles, as these parked vehicles act as temporary spatial obstructions. Furthermore, since parked vehicles are dynamic and may relocate over time, spatial information obtained from satellite maps or other predefined sources may not accurately reflect real-time road conditions, leading to erroneous sensor interpretations. To address this limitation, we propose an NLoS pedestrian localization framework that integrates monocular camera image with 2D radar point cloud (PCD) data. The proposed method initially detects parked vehicles through image segmentation, estimates depth to infer approximate spatial characteristics, and subsequently refines this information using 2D radar PCD to achieve precise spatial inference. Experimental evaluations conducted in real-world urban road environments demonstrate that the proposed approach enhances early pedestrian detection and contributes to improved road safety.
A novel radar-based non-line-of-sight (NLoS) pedestrian localization framework that enables robust detection of darting-out pedestrians near parked vehicles by integrating radar signals with camera-assisted point cloud interpretation.
System Overview: The proposed system leverages radar and camera-based depth perception to enable Non-Line-of-Sight (NLoS) pedestrian localization in darting-out scenarios near parked vehicles. First, the camera assistant module processes a front camera image to perform vehicle segmentation and generate a depth point cloud (PCD), capturing spatial depth information. Next, the spatial inference module integrates the vehicle depth PCD with radar PCD, distinguishing between static and moving points while refining vehicle predictions. Finally, the NLoS object localization module processes the radar data to detect pedestrians hidden behind occlusions and localizes their positions. The final results are validated against BEV ground truth, demonstrating the system’s effectiveness in real-world urban environments.
We designed three experimental scenarios, SA, SB, and SC, to evaluate the effectiveness of our proposed NLoS pedestrian localization system in darting-out situations near parked vehicles. In SA, a single pedestrian suddenly appears from behind a parked vehicle, testing the system’s ability to detect and track a pedestrian in a fully NLoS setting where direct visibility is obstructed. SB involves two pedestrians emerging sequentially from behind a parked vehicle, where the first pedestrian acts as an additional occlusion for the second pedestrian. This scenario evaluates the system’s ability to detect a pedestrian under dynamically changing occlusion conditions. SC consists of one pedestrian walking along a clear Line-of-Sight (LOS) path while another pedestrian emerges from the NLoS region, demonstrating the system's capability to simultaneously detect pedestrians in both LOS and NLoS conditions. These scenarios assess the system’s robustness in handling various occlusions caused by parked vehicles and moving pedestrians.
SA
SB
SC
