Multimodal Bayesian Filtering for Enhanced Road Surface Condition Classification in Intelligent Vehicles Safety Systems

초록

This paper presents a multimodal Bayesian filtering for reliable road surface condition classification. The proposed framework integrates a vision-based classifier and a data-driven gated recurrent unit model that uses vehicle sensor data. By probabilistically fusing the outputs of both models through a Bayesian filter, the method effectively compensates for the weaknesses of single-modality. Experimental results demonstrate that the proposed fusion algorithm achieves higher accuracy and robustness under various environmental and road conditions compared to conventional vision- or sensor-based methods. These findings highlight the potential of multimodal Bayesian fusion for enhancing road perception reliability in intelligent vehicle systems. © 2026 IEEE.

키워드