Abstract
To enhance road safety and prevent traffic accidents caused by overlimit trucks, this study developed a truck dimension detection technology based on LiDAR-equipped drones. Drones, with their high mobility, rapid deployment capability, and adaptability to complex road environments, serve as an ideal platform for such tasks addressing the limitations of traditional fixed-target methods and computer vision approaches. Compared with fixed-target LiDAR systems that typically have a measurement error of over 5%, this technology achieves significantly higher precision. A quadrotor UAV was used as the measurement platform, integrated with a high-precision mechanical LiDAR system to construct a hardware setup capable of real-time vehicle dimension measurement. Wireless modules were incorporated to enable real-time remote data transmission. To optimize the point cloud data captured by LiDAR, this study employed a K-D tree algorithm and combined it with Principal Component Analysis (PCA) to propose an efficient method for vehicle dimension detection. This method differs from existing UAV-based detection studies: a relevant study adopts a deep learning approach that focuses on abnormal state recognition, whereas the PCA-based bounding box algorithm in this study specifically enhances the extraction of vehicle contour features from noisy point clouds (via K-D tree preprocessing), thereby reducing dimension measurement deviations. When the drone is in a hovering state, the measurement error of this method is less than 2.0%. The research outcomes not only enhance the monitoring capability of overlimit transport vehicles but also provide traffic management authorities with an efficient and convenient dynamic monitoring tool. This technology can play a significant role in reducing traffic accident risks and optimizing road safety management.
Get full access to this article
View all access options for this article.