A step S1 of inputting coordinate values on a three-dimensional shape in a new measurement position, a step S4 of dividing a spatial region where the three-dimensional shape is present into voxels composed of cubes, and making an environment model storing the voxel positions, a matching step S5 of setting/storing a representative point in the voxel corresponding to each coordinate value and the error distribution, and a fine aligning step S7 of rotating and translating the new measured data and the error distribution of the environment model corresponding to the previously measured position or rotating and translating the environment model corresponding to the new measured position, and performing alignment, if any data is present at the previous measurement position, so that the total ofthe distances between error distributions near to each other may take on a minimum value. From the amount of rotation and the amount of translation at the fine aligning step, the self-position is identified.