A Novel Point Cloud Clustering Algorithm Integrating In-Vehicle Driver Monitoring

Real-time, high-precision point cloud clustering algorithm is the prerequisite to achieve accurate environment perception and scene understanding in human-machine co-driving systems. This paper presents a novel point cloud clustering algorithm integrating in-vehicle driver monitoring. The algorithm initially employs LiDAR for the collection of point cloud information. And it utilizes a camera to continuously monitor the driver's head position and the direction of their eye gaze in real-time. The algorithm then utilizes the improved TOPSIS multi-criteria decision-making approaches to determine the optimal spatial segmentation of the point cloud of interest, followed by applying a scanline clustering algorithm based on Euclidean distance and vector fusion for point cloud clustering. Experiments in real-world environments demonstrate that the proposed clustering algorithm can significantly enhance both accuracy and computational speed in comparison to current mainstream clustering algorithms. These improvements are expected to enhance the performance of future human-machine co-driving systems.