The invention discloses a method for automatically extracting building information in airborne laser radar point clouds. According to the invention, an airborne laser radar system is employed to carry out data acquisition on a target zone, three point clouds which are not coplanar are randomly selected from original point clouds according to a constraint condition, the equation of a plane determined by the selected point clouds is calculated in dependence on space coordinates of the selected point clouds, the distance di between each point cloud and the plane is calculated, the standard deviation [sigma] of the distance is calculated, and the threshold t is set to satisfy the formula t=2 [sigma]. When di is larger than t, the point cloud is classified as an outside point, conversely, the point cloud is classified as an inside point, and the number of the inside points is counted up; the above steps are iterated for certain times, a plane including the most inside points is selected, the point clouds in the obtained plane model are removed, residual points serve as an original point cloud data set for next processing, the above steps are repeated until all point cloud planes of the building are extracted, and finally a part of misclassified point clouds are removed by means of three-dimensional morphological corrosion operation. According to the invention, the building point cloud extraction speed and precision are improved.