Plane detection in 3-D space is a core function of the autonomous mobile robot. A representative technique for plane detection is the Hough transform method. The Hough transform is robust to noise and makes accurate plane detection possible. However, a common problem in methods based on the Hough transform is that too much time is required to calculate parameters, which adds computational cost and memory requirements for parameter voting to find the distribution of mixed multiple planes in the parameter space. Furthermore, real-time processing for sequential image sequences is challenging, because the whole process must be repetitively performed for the next detection. We extend the conventional self-organizing map by introducing a real-time clustering method and by detecting multiple planes through the creation, extinction, renewal, and merging of plane parameter data, which are input sequentially. The proposed method is also based on reliable plane detection through a planarity evaluation during data sampling. The results of experiments conducted under various conditions with an unmanned vehicle demonstrate that the proposed method is more accurate and faster than conventional methods.