Abstract: Driverless technology in mines is becoming a research hotspot in recent years, and one of the core links is the road passable area recognition method for path planning. The open-pit mine road scene is characterized by ambiguous road boundaries and uneven road smoothness, etc. However, the existing ground segmentation algorithms are prone to misclassification problems, such as the ground points are not connected to the vehicles and the ground segmentation results are inconsistent between frames. Therefore, this paper proposes a spatio-temporal continuity-based approach to recognize the passable area of mine roads. The method firstly models the mine road based on concentric circle model and uses principal component analysis to obtain the initial passable area segmentation results; then, based on spatial connectivity, the local region generation algorithm and K-nearest neighbor algorithm are respectively used to perform region-level connectivity filtering and point-level connectivity filtering for the initial passable area, and the passable area in line with spatial connectivity is obtained based on the results of the two-step filtering. Finally, the unstable regions with inconsistent traversability in different point cloud frames are filtered out based on the temporal region consistency, the raster map is constructed according to the normal distribution transformation algorithm, and then the distance weight formula is used to judge the stability of the raster, and finally the unstable regions are filtered out by the regional raster projection. The experimental results show that the precision rate of the passable region recognition algorithm described in this paper is 93.44%, which is 2.27% higher than the existing algorithm, and the recall rate of the algorithm is 99.14%, which is 8.26% higher than the existing algorithm.