The project "YOLO-UNET-based visual road guidance system" aims to provide accurate and efficient road and traffic sign recognition for unmanned and intelligent transportation systems. The project integrates YOLOv5-7.0 models with UNET technology and is trained and deployed with Pytorch and TensorRT. CUDA technology ensures high-speed acceleration of computing. Using UNET and OpenCV processing, we get a high-contrast binary image, optimize the line inspection function, and realize the accurate unilateral line inspection function through curve fitting. This system shows innovation and technological excellence, conforms to the concept of "intelligent change of life, creative improvement of quality", and expects to promote the development of intelligent transportation, and provide safety and efficiency guarantee for the future transportation system.