In this chapter we consider the detection of hazards within the ground plane immediately in front to a moving pedestrian. Consecutive views of the scene are acquired by a standard video camera. Using epipolar constraints between the two views, detected features are matched to compute the camera motion and reconstruct the 3-D geometry. Assuming the ground is planar, projective invariance of the cross-ratio and the presence or absence of significant peaks in a Lomb-Scargle periodogram are used in a region-growing technique to label a triangulated mesh as obstructed or unobstructed ground plane. On the other hand, for a less feature based scene a new disparity velocity based obstacle detection scheme is presented. This scheme can be used to find image points of large disparity estimates and hence single out suspicious obstructed ground points. The experimental work shows the performance of these two algorithms in real image sequences.