Building modeling method using aerial lidar and computer program recorded on recording medium to execute the same

1 . A building modeling method, comprising:

separating, by a data generating device, a point cloud corresponding to a ground and a plurality of point clouds corresponding to a non-ground from aerial point cloud data acquired from a LiDAR mounted on a flight device;

classifying, by the data generating device, a point cloud corresponding to a roof of a building among the plurality of point clouds corresponding to the non-ground; and

modeling, by the data generating device, the building based on the classified point cloud,

wherein, in the modeling, an edge of the roof is estimated based on a distribution of the classified point cloud, and an outer point of the classified point cloud is fit based on the estimated edge, and

wherein, in the modeling, a covariance matrix using neighboring points of each point included in the classified point cloud is obtained, an eigenvector is obtained through principal component analysis (PCA) on the covariance matrix, and two eigenvectors in which an angle therebetween is within a threshold value are estimated as the edge of the roof.

2 . The building modeling method of claim 1 , wherein,

in the separating the point cloud,

the aerial point cloud data is voxelized to a preset size, and other points, except for a point closest to a center point, among points in a voxel, are deleted to acquire a uniform sample.

3 . The building modeling method of claim 2 , wherein, in the separating the point cloud, the aerial point cloud data is divided into a grid smaller than an average distance between points to generate row and column indexes for each point, a height value of the grid is set using an interpolation method when there is no point in the grid, an opening calculation is performed using a window having a preset size, cases before and after the opening calculation is performed are compared based on a height threshold value to separate the point cloud corresponding to the ground and the point cloud corresponding to the non-ground from each other.

4 . The building modeling method of claim 1 , wherein,

in the classifying the point cloud,

the point cloud corresponding to the roof of the building and a point cloud corresponding to a tree among the plurality of point clouds corresponding to the non-ground are classified based on a number of reflected signals reflected from one pulse from the LiDAR.

5 . The building modeling method of claim 4 , wherein,

in the classifying the point cloud,

the point cloud in which the number of reflected signals reflected from the one pulse of the LiDAR exceeds a preset value is identified as the tree.

6 . The building modeling method of claim 1 , wherein,

in the classifying the point cloud,

a point corresponding to a reflected signal having a signal strength smaller than a preset value, among reflected signals reflected from one pulse of the LiDAR is deleted.

7 . The building modeling method of claim 1 , wherein,

in the modeling,

a basic primitive of the classified point cloud is detected, and the classified point cloud is modeled as the roof according to the detected basic primitive.

8 . A non-transitory computer-readable recording medium having stored thereon a computer program that, when executed by a processor, causes the processor to execute operations, comprising:

separating, by the processor, a point cloud corresponding to a ground and a plurality of point clouds corresponding to a non-ground from aerial point cloud data acquired from a LiDAR mounted on a flight device;

classifying, by the processor, a point cloud corresponding to a roof of a building among the plurality of point clouds corresponding to the non-ground; and

modeling, by the processor, the building based on the classified point cloud,

wherein, in the modeling, an edge of the roof is estimated based on a distribution of the classified point cloud, and an outer point of the classified point cloud is fit based on the estimated edge, and

wherein, in the modeling, a covariance matrix using neighboring points of each point included in the classified point cloud is obtained, an eigenvector is obtained through principal component analysis (PCA) on the covariance matrix, and two eigenvectors in which an angle therebetween is within a threshold value are estimated as the edge of the roof.

9 . A building modeling method, comprising:

separating, by a data generating device, a point cloud corresponding to a ground and a plurality of point clouds corresponding to a non-ground from aerial point cloud data acquired from a LiDAR mounted on a flight device;

classifying, by the data generating device, a point cloud corresponding to a roof of a building among the plurality of point clouds corresponding to the non-ground; and

modeling, by the data generating device, the building based on the classified point cloud,

wherein, in the separating the point cloud, the aerial point cloud data is voxelized to a preset size, and other points, except for a point closest to a center point, among points in a voxel, are deleted to acquire a uniform sample, and

wherein, in the separating the point cloud, the aerial point cloud data is divided into a grid smaller than an average distance between points to generate row and column indexes for each point, a height value of the grid is set using an interpolation method when there is no point in the grid, an opening calculation is performed using a window having a preset size, cases before and after the opening calculation is performed are compared based on a height threshold value to separate the point cloud corresponding to the ground and the plurality of point clouds corresponding to the non-ground from each other.