A Study on Point Cloud Clustering for Geo-referenced Positioning using mmWave Radar Sensor

In the field of urban safety, extensive research is being conducted on the application of mmWave radar sensors for both indoor and outdoor environments. These advanced sensing technologies are gaining attention due to their potential to enhance monitoring and response capabilities in complex urban settings. Traditional CCTV-based geo-referenced positioning technology predicts Geographic coordinates data to enable real-time site response in the event of an incident. However, pixel-based relative geo-referenced methods of CCTV suffer from low recognition rates and precision due to terrain characteristics. mmWave radar sensors, in contrast, are minimally affected by weather conditions, use high frequencies in the 60-63 GHz range, and offer high precision. Additionally, these sensors can penetrate structures up to 15 cm thick. Despite these advantages, the number, shape, and position of recognized objects in three-dimensional space vary depending on the sensor installation angle and position, complicating the application of geo-referenced technology for accurate positioning. Consequently, mmWave radar sensors have not been widely utilized for positioning. This study proposed and implemented a geo-referenced positioning model using mmWave radar sensors to address these technical challenges. The proposed model employed a dimension transformation clustering technique based on orthogonal projection methods. Implementation and experimental results demonstrated that the average georeferenced error using mmWave radar sensors was reduced to 3.5 cm, representing a $95.33 \%$ improvement in precision compared to traditional CCTV-based geo-referenced methods.