AlgorithmDescription

MCC Algorithm

SD : scale domain, integer in [1,3] (variable l (lowercase "L") in paper)

t_SD: curvature tolerance for scale domain SD
t₁ = specified by user
t₂ = t₁ + 0.1
t₃ = t₂ + 0.1

CR_SD : cell resolution for scale domain SD (lambda variable in paper)
CR₁ = 0.5 * CR₂
CR₂ = specified by user (nominal post-spacing of input LiDAR data)
CR₃ = 1.5 * CR₂

f : tension parameter (invariant across scale domains)
f = 1.5

U : vector of points that remain unclassified
U = (P₁ , P₂ , P₃ , ... P_n)
n : # of points in U (at the start of each loop pass)
P_j : single LiDAR point
P_j = (x_j , y_j , z_j)
U₀ = initial point-cloud data specified by user

1. U = filter out higher points at same x,y in U₀ and classify them as non-ground
2. for scale domain (SD) = 1 to 3:
3. . . repeat
4. . . . . S = interpolate new raster surface using TPS(U, CR_SD, f)
5. . . . . S' = surface resulting from passing 3x3 averaging kernel over S
6. . . . . for each point P_j in U:
7. . . . . . . if z_j > S'(x_j , y_j) + t_SD then
8. . . . . . . . . classify P_j as non-ground and remove it from U
9. . . . . n_C : # of points classified and removed from U during current iteration through inner loop
10. . . until n_C < 0.1% * n
11. classify all the points remaining in U as ground