The detection of dynamic points on a moving platform is an important task to avoid a potential collision. However, it is difﬁcult to detect dynamic points using only two frames, especially when various input data such as ego-motion, disparity map, and optical ﬂow are noisy for computing the motions of points. In this paper, we propose a supervised learningbased approach to detect dynamic points in consideration of noisy input data. First of all, to consider depth ambiguity that proportionally increases according to the distance to the egovehicle, we divide the XZ-plane (bird-eye view) into several subregions. Then, we train a random forest for each subregion by constructing motion vectors computed based on two motion metrics. Here, in order to reduce errors of the input data, the motion vectors are ﬁltered based on a pairwise planarity check and then ﬁltered motion vectors are used for training. In the experiments, the proposed method is veriﬁed by comparing the detection performance with that of previous approaches on the KITTI dataset.