Identifying and removing bad lines in a shapefile

Question

I downloaded some shapefiles of Houston from the Open Maps data using this tool. I am plotting the roads and I have a weird thing happening. Most of the lines are correct, but there are clearly several erroneous lines that are drawing long streaks across the map.

How can I identify these lines and remove them from the shapefile?

Here is the code I used to plot the thing.

import geopandas
import matplotlib.pyplot as plt

fname = 'planet_-96.09,29.269_-94.616,30.18-shp/shape/roads.shp'

fig = plt.figure(dpi=300)
ax = plt.subplot(1, 1, 1)

lat_0 = -95.579
lat_1 = -94.99
lon_0 = 29.577
lon_1 = 30.19

ax.set_xlim([lat_0, lat_1])
ax.set_ylim([lon_0, lon_1])

roads = geopandas.read_file(fname)

roads.plot(ax=ax, lw=.05)

plt.savefig('houston.png')

Answer 1

Based on the plot you have shared, it looks like the problematic lines you want to get rid of are the ones that are straight containing just two coordinates. If this assumption is correct, you could remove them from your geodataframe following these steps:

First, create a function that counts the number of coordinates or vertices for each line. Something like this should do the trick:

  def get_num_vertices(row):
    if row.geometry.type.startswith("Multi"):
        num_vertices = 0
        for part in row.geometry:
          num_vertices += len(part.coords)      
    else:
      num_vertices = len(row.geometry.coords)
    return num_vertices

Secondly, drop all rows with less than two coordinates/vertices (and any other filter like geometry.length greater than...).

roads['num_vertices'] = roads.apply(lambda x: get_num_vertices(x), axis=1)
roads_filtered = roads.loc[roads.num_vertices > 2]