Abstract: Point cloud registration is the important part in the large-scale city reconstruction. Considering the ground plane and buildings, which are the main objects in the city, always consist of several planes, we propose a new method that registering large-scale point clouds together through matched planes. For extracting the planes exactly from the point cloud, this paper presents a novel co-clustering method that clusters data points and hypotheses simultaneously. The traditional methods only cluster the data points, so they will fail to extract some planes. After knowing the planes, we use the random sample method to match the planes. The transformation matrix between the point clouds is calculated from the matched planes. Experimental results not only show that the co-clustering method can extract the planar structure more accurate and robust, but also demonstrate the effectiveness of the novel registration method for the city point clouds. Since the process in our method is only performed to the sparse version of the point clouds, it reduces the amount and complexity of calculation, which makes it to be suitable for the large-scale city reconstruction.