Abstract: Improving the environment perception ability is a prerequisite for improving the intelligence and autonomy of the rover. Non-geometric hazard perception is the key part of autonomous navigation. Reviewing the development of the non-geometric hazard perception of the rover, it shows that the key aspects of the non-geometric hazard perception are hazard estimation and hazard prediction. Non-geometric hazard estimation and prediction include wheel-soil model, wheel sinkage estimation, slip ratio estimation, identification of soil parameters and terrain classification, and the relevant research on theses aspects are introduced and analyzed. The future research on non-geometric hazard perception should focus on the optimization of wheel-soil interaction mechanics model, multi-source information fusion and the application of artificial intelligence technology. Finally, a slip ratio prediction method based on soil parameter identification is described.