Next-Gen Pedestrian Monitoring: Advanced LiDAR and IoT Solutions for Urban Health and Safety

A study conducted by researchers from Missouri University of Science and Technology, USA, and King Abdullah University of Science and Technology, Thuwal, Saudi Arabia explores how advanced technologies, specifically 3D LiDAR sensors and the Internet of Things (IoT), can be used to improve urban safety and public health. It proposes a new framework for monitoring pedestrian activities and detecting abnormal behavior in city traffic.

Three-Phase Methodology for Pedestrian Monitoring

The proposed methodology consists of three main phases. The first phase is data collection and labeling. LiDAR data is collected to capture 3D aspects of traffic situations, and this data is meticulously labeled for vehicles and pedestrians. The second phase involves 3D object detection using a modified Point Voxel-Region-based Convolutional Neural Network (PV-RCNN). This model processes the labeled data to identify vehicles and pedestrians, significantly reducing the risk of misclassification. The third phase is pedestrian activity classification. Point clouds from identified pedestrians are classified as either 'Normal' or 'Abnormal' using PointNet, a model that helps in understanding pedestrian behaviors and enhancing urban safety.

Elevated LiDAR Sensors and Data Collection

Elevated LiDAR sensors are strategically positioned on urban infrastructure, such as traffic lights and street lamps, to monitor pedestrian activities. This setup captures detailed 3D point cloud data, providing comprehensive insights into pedestrian behavior. The collected data helps distinguish between normal activities, like walking, and abnormal activities, like walking with an injured leg. This distinction is crucial for developing targeted public health interventions designed to improve pedestrian safety and health. To overcome the challenges of collecting real-world data, especially for risky scenarios like falls, the researchers used the Blender simulator to create detailed traffic scenes. These scenes depict various pedestrian activities, both normal and abnormal, and include vehicles. LiDAR sensors placed in the simulated environment capture 3D coordinates and reflectivity values, creating a robust dataset for analysis.

Object Detection and Activity Classification

The PV-RCNN model, which has been fine-tuned for this study, processes the raw 3D point cloud data to detect objects. The model uses attributes like 3D coordinates and intensity to generate accurate bounding boxes for pedestrians and vehicles, enhancing detection precision. The hybrid approach of PV-RCNN, combining point and voxel features, ensures detailed and comprehensive 3D object detection. After detecting objects, the next step is to classify pedestrian activities using PointNet. This model processes the 3D point cloud data to classify pedestrian activities, distinguishing between normal and abnormal behaviors. The classification process is fine-tuned to ensure high accuracy, focusing on both local and global features of the point cloud data.

Promising Results and Impact on Urban Safety and Public Health

The dataset includes 21 diverse urban scenarios simulated in Blender, capturing a range of pedestrian activities. Metrics used to evaluate the performance of the methodology include Average Precision (AP), Recall, Precision, and F1-Score. The PV-RCNN model outperforms the SECOND model in detecting both pedestrians and vehicles, achieving higher metrics and demonstrating its superior capability in leveraging 3D point cloud data for accurate object detection. The PointNet model shows remarkable accuracy in classifying pedestrian activities, effectively distinguishing between normal and abnormal behaviors with high precision and recall rates. This performance highlights the model's potential to enhance urban safety through detailed behavioral analysis.

The study presents a comprehensive framework using 3D LiDAR-based point cloud data to monitor and classify pedestrian activities. By integrating advanced technologies, the framework significantly improves urban safety and public health monitoring. The research contributes to the field by providing detailed insights into pedestrian behavior, promoting safer urban environments, and enhancing public health interventions. This innovative approach not only helps in better managing urban traffic but also in developing more effective public health strategies, ultimately leading to safer and healthier cities.