alessandro-gentilini/Efficient-RANSAC-for-Point-Cloud-Shape-Detection

Numerical example of sampling effect in the source paper

guillermoacdc opened this issue · 0 comments

Has someone confirmed the values ​​of the numerical example presented in section 4.3.1 of the source paper [1]?

In my attempt I obtained different values; (assumption: d=2)
Expected Values:

T = 151.522.829
T_local = 64.929

Obtained Values

T = 1.6210e+07
T_local = 3.4558e+04

[1] R. Schnabel, R. Wahl, and R. Klein, “Efficient RANSAC for point-cloud shape detection,” Eurographics Assoc. Blackwell Publ. 2007, vol. 26, no. 2, pp. 214–226, 2007.

my code in matlab
%% parameters
pt=0.99;%probabilitiy of a succesful detection; defined by user
k=2;%size of a minimal set required to define a cylinder candidate.
n=1066;%size of the shape psi in the point cloud PC
N=2000000;%size of the point cloud PC; in number of points
d=2;%--octree's depth where the cell C is located; Points in cell=341.547

%% compute P(n)
Pn=(n/N)^k;%probability of detecting the shape psi in a single pass
Pn_local=n/(Nd2^(k-1));

%% compute T
T=log(1-pt)/log(1-Pn)
T_local=log(1-pt)/log(1-Pn_local)