NaN eigentropy

[patcmorneau]

input of line 83 and 84 after sort | uniq of original xyz file :

83	594904.5999 5461748.8828 -11.0322
84	594904.6026 5461748.8829 -11.0322

output sample of sounding generate feature:

1 0 0
line 83 contain NaN
594877.125000,5461807.500000,-9.342700,0.000292615181,0.000000000000,nan,1.000000000000,0.000000000000,1.000000000000,0.000000000000,0.000000000000,1.000000000000,-0.000000000000,0.000000000000,0.002633536629,0.000000000000,0.057000160217,-0.032500267029,0.024499893188

1 0 0
line 84 contain NaN
594877.125000,5461807.000000,-9.480900,0.000220370923,0.000000000000,nan,1.000000000000,0.000000000000,1.000000000000,0.000000000000,0.000000000000,1.000000000000,0.000000000000,0.000000000000,0.001983338306,0.000000000000,0.044400215149,-0.044400215149,0.000000000000

output explanation :
line 1 : the three eigen value
line 2 : nan
line 3 : xyz features, 6th feature cause nan because of log(eigen value = 0)

problem:
points are too close and the roundup makes them the same wich result into no variation and therefore a logarithm error

[patcmorneau]

using double instead of float to remove the roundup error is not possible since the library PCL only support floats for the class pointXYZ.
line 76 : cloud->push_back(pcl::PointXYZ(valueX,valueY,valueZ) );