Efficient Pure Python 3 implementation of Visvalingam and Whyatt's algorithm for reducing the complexity of poly-lines. Also includes precision decimation to reduce file sizes if desired.
Works with GDAL OGRGeometry LINESTRING, POLYGON and MULTIPOLYGON objects as well as lists of vertices.
This method ranks the vertices by their importance to the shape (how much removing the affects the area) in a non-destructive manner. Once the ranking has been done, filtering of points is ultra fast (just one numpy mask operation).
However, even for just one filtering operation, this method seems to be faster than Ramer-Douglas-Peucker.
from polysimplify import VWSimplifier
import numpy as np
from time import time
n = 5000
thetas = np.linspace(0, 2 * np.pi, n)
pts = np.array([[np.sin(x), np.cos(x)] for x in thetas])
start = time()
simplifier = VWSimplifier(pts)
VWpts = simplifier.from_number(n / 100)
end = time()
print("Visvalingam: reduced to %s points in %03f seconds" % (len(VWpts), end - start))
# 50 points in .131 seconds on my computer
from rdp import rdp
start = time()
RDPpts = rdp(pts, epsilon=.00485) # found by trail and error
end = time()
print("Ramer-Douglas-Peucker: to %s points in %023 seconds" % (len(RDPpts), end - start))
# 40 points in 1.35 seconds on my computer