The lane line is difficult to be distinguished in visible light for driver assistance, so a multi-feature fusion model is built from the perspective of infrared images to realize assistant driving. Firstly, the features of infrared imaging are analyzed, and the Enet network is improved to focus on the area of the lane line and locate the passable area. Then, the previous vehicle is located to guide the current vehicle based on the fuzzy set theory. Finally, a spatiotemporal association model is constructed. By constructing the relationship between the guiding vehicle and spatiotemporal traffic, the relationship between human-computer interaction is indirectly established to guide vehicle assisted driving. Our experimental results show that the proposed algorithm is in line with the manual driving process, and good results can be achieved under complex road conditions. © 2021