In the three-dimensional point cloud registration, when the surface of the point cloud is relatively flat and the geometric features are fuzzy, the iterative closest point algorithm has poor registration results, even often fails to register. The point cloud data obtained by the three-dimensional laser scanner includes geometric coordinate information and RGB information. Here, by making full use of point cloud coordinate information and RGB information, we propose a new point cloud registration method, which first convert RGB values into grayscale values, set the weighting factor according to the sum of the variance of the gray value and the sum of the variances of each curvature, and then adaptively adjust the impact of color information and geometric information on registration in the light of the weighting factor to achieve an organic combination based on color information and geometric information. Experimental results show that the proposed method can achieve stable and accurate registration of different point clouds.