In the point cloud splicing of underground pipe gallery in power plants

an improved algorithm with 3DHarris operator about corner detection is proposed. In this algorithm

firstly

taking multiple sets of massive point cloud data of the underground pipe gallery as analysis object

the normal information of the target detection point on the adjacent point cloud microtangent plane is obtained through principal component analysis method

thus the boundary points of the point cloud are extracted. Secondly

the covariance matrix is constructed according to the normal information of the target detection point

then the corner response intensity function is calculated and compared

and some points are selected as calculation objects to judge whether they are true corner. Thirdly

the pseudo angles are filtered out and the true angles are selected by using a pseudoangle detection method based on Gaussian curvature extreme points. At last

the fast point feature histogram method is used to match similar corners between each group

and the nearest point search point cloud registration algorithm is used to realize the splicing of multiple points and cloud in the underground pipe gallery. The results are compared with that of the conventional 3D-Harris angular point detection algorithm. It is proved that the method in paper takes less time to calculate and has a higher corner-point extraction accuracy

which can realize accurate splicing of massive point cloud of underground pipe gallery.