Vehicle detection based on fusion of millimeter-wave radar and monocular vision

In the realm of intelligent driving, mmWave radar and camera sensors hold paramount importance. Their synergistic integration not only broadens the scope of perception but also enhances accuracy, which is pivotal for collaborative Vehicle-to-Everything (V2X) communication. This article delves into the precision of fusion between mmWave radar and camera data. Initially, within the processing of mmWave radar, we employ the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) clustering algorithm for preprocessing radar data and establish connections with corresponding and adjacent frame data. In the realm of visual processing, we harness YOLOv5 (You Only Look Once version 5), for object detection. Subsequently, under spatiotemporal alignment, we match relevant targets between mmWave radar and imagery and fuse them using a weighted approach. To further bolster accuracy, the Extended Kalman Filter (EKF) tracks both matched and unmatched targets. Experimental results demonstrate that this method elevates target detection precision by effectively amalgamating data from both sensors.