The background of the improvement of multi-sensor fusion technology in intelligent autonomous driving is to improve the vehicle’s perception of complex environment and ensure the safety and reliability of driving. In today’s society, the accuracy and efficiency of multi-sensor fusion technology in intelligent autonomous driving are not high, which greatly affects the safety of multi-sensor fusion technology in intelligent autonomous driving. The YOLOv5 algorithm in machine learning is an effective multi-sensor fusion technology in intelligent autonomous driving technology. YOLOv5 algorithm is used to build multi-sensor fusion technology in intelligent autonomous driving system, which power the accuracy and safety of multi-sensor fusion technology in intelligent autonomous driving. Finally, the experimental results show that the YOLOv5 algorithm is easy to operate and the accuracy rate reaches 99.9%.