class RobustSurfaceSplitting:
def __init__(self, point_cloud, initial_line_points_file):
"""
point_cloud: The set of points that needs to be split
Point cloud contains the points that needs to be split
Initially the front curved surface and the planar wall to its left are considered(For testing)
"""
self.pointcloud = point_cloud
self.lineFitting = LineFitting(np.matrix(np.genfromtxt(initial_line_points_file)))
self.line_params, initial_error = self.lineFitting.get_lineparameters() # [theta0, theta1] shape 2,1
print "initial error", initial_error
self.Q = 50 # Q is the number of points close to the splitting line
self.number_iterations = 1
def __splitPointCloud(self):
"""
the point cloud is split into two according to the sign of l(X)
"""
x = self.pointcloud[:, 0]
y = self.pointcloud[:, 1]
ones = np.matrix(np.ones(x.shape[0])).T
values = (np.hstack((x, ones)) * self.line_params) - y
# Return two sets of points based on the sign
values = np.array(values).ravel()
P1 = np.where(values < 0)[0]
P2 = np.where(values > 0)[0]
return P1, P2
def split(self):
# Step 1: Split the points in the points cloud into P1 and P2
# Step 2: Fit appropriate surfaces to P1 and P2
# Step 3: Find the Distance D = | S1(x, y) - S2(x, y) |
# Step 4: Find the subset of Points Q that have least distance
# Step 5: Fit a line to this set of points. Repeat
for i in xrange(self.number_iterations):
self.p1_indices, self.p2_indices = self.__splitPointCloud()
self.p1, self.p2 = self.pointcloud[self.p1_indices], self.pointcloud[self.p2_indices]
# Fit new surfaces to the line
P1 = np.array(self.p1)
P2 = np.array(self.p2)
self.linearSurfaceParameters, self.DLinearSurface, self.indicesLinearSurface = LinSurfFit(
P1[:, 0], P1[:, 1], P1[:, 2]
)
chosenPointsLinear = self.p1[self.indicesLinearSurface]
chosenPointsLinear = get_new_surface(chosenPointsLinear, self.DLinearSurface, self.linearSurfaceParameters)
self.quadraticSurfaceParameters, self.DQuadratic, self.indicesQuadraticSurface = QuadSurfFit(
P2[:, 0], P2[:, 1], P2[:, 2]
)
chosenPointsQuadratic = self.p2[self.indicesQuadraticSurface]
chosenPointsQuadratic = get_new_surface(
chosenPointsQuadratic, self.DQuadratic, self.quadraticSurfaceParameters
)
write_ply_file(chosenPointsLinear, "../plyfiles/iteration" + str(i + 1) + "LinearSurface.ply")
write_ply_file(chosenPointsQuadratic, "../plyfiles/iteration" + str(i + 1) + "QuadraticSurface.ply")
write_ply_file(chosenPointsLinear, "../front_left_surface/left_surface.ply")
write_ply_file(chosenPointsQuadratic, "../front_left_surface/front_surface.ply")
self.pointcloud = np.vstack((chosenPointsLinear, chosenPointsQuadratic))
write_ply_file(self.pointcloud, "../plyfiles/iteration" + str(i + 1) + "newpointcloud.ply")
def __init__(self, point_cloud, initial_line_points_file):
"""
point_cloud: The set of points that needs to be split
Point cloud contains the points that needs to be split
Initially the front curved surface and the planar wall to its left are considered(For testing)
"""
self.pointcloud = point_cloud
self.lineFitting = LineFitting(np.matrix(np.genfromtxt(initial_line_points_file)))
self.line_params, initial_error = self.lineFitting.get_lineparameters() # [theta0, theta1] shape 2,1
print "initial error", initial_error
self.Q = 50 # Q is the number of points close to the splitting line
self.number_iterations = 1
