Multi-source data fusion method based on nearest neighbor plot and track data association

In response to the multi-sensor data fusion about intelligent vehicle, an improved algorithm for vehicle based on plot and track data association is proposed. This method takes Euclidean distance and cosine similarity as well as target track as indexes and sets up a reasonable maximum threshold for tracking based on the target history state. Then, the targets identified from the visual system and radar system are fused by using the probability distribution to achieve the maximum utilization of the detection range of sensors. On-vehicle experiments are executed to evaluate the performance of this proposed method, and results demonstrate that through the comparison of fusion data and single sensor data, this proposed algorithm integrates the advantages of fusion system to accurately perceive the environmental information.