rasterlinefilter - classify lines by underlying raster

Contents

• Overview
• Usage
• Example
• Notes

Overview

rasterlinefilter.py reads lines and samples classes in a raster under the lines, producing new line segments with the classes from the raster. Classes can be lumped without needing to reclassify the raster, e.g. for a raster with classes A,B,C,D,E,F rasterlinefilter can generate class N line segments for raster classes A,B,C and class O line segments for raster classes D,E,F.

A threshold number of consecutive samples within a (lumped) raster class before the line classification changes can be specified. This allows output line segments to bridge over fine scale / noisy raster data and represent major features in the raster.

Vertices present in the original input line data are always present in the output, even when they don't represent a class change. This is necessary to maintain line shape. Sampling vertices are only present in the output when they occur at class change boundaries, i.e. new line segments.

Pictorially:

original line with vertices

X--------------------------X----------X--------------------X---------------X

line with additional sampling vertices

X----x----x----x----x----x-X----x-----X----x----x----x-----X----x----x-----X

raw raster classes

AAECBBEACCAADDBFFEECDDCAAFBBBACBCAEEDEEDDEEEEAEEEABBABDFCFFAAAEBCCDDCEEEABFF

lumped raster classes (A,B,C = N, D,E,F = O)

NNONNNONNNNNOONOOOONOONNNONNNNNNNNOOOOOOOOOOONOOONNNNNOONOONNNONNNOONOOONNOO

output six separate line segments with class

X---------x---------x------X----x-----X---------x----------X----x----------X
NNNNNNNNNNNOOOOOOOOOONNNNNNNNNNNNOOOOOOOOOOOOOOOONNNNNNNNNNNNNNNNOOOOOOOOOOO

Usage

usage: rasterlinefilter.py [-h] [--step-length STEP_LENGTH]
                           [--stretch STRETCH] [--min-steps MIN_STEPS]
                           [--class-steps CLASS_STEPS] [--class CLASS_]
                           [--values VALUES] [--fields FIELDS [FIELDS ...]]
                           [--range RANGE] [--band BAND] [--get-classes]
                           [--progress PROGRESS]
                           lines grid output

Clasify lines by rasters

positional arguments:
  lines                 Path to OGR datasource (shapefile) containing lines
  grid                  Path to GDAL datasource (grid) containing raster
  output                Basename for output shapefile

optional arguments:
  -h, --help            show this help message and exit
  --step-length STEP_LENGTH
                        Spacing of sampling nodes along lines (default: 10.0)
  --stretch STRETCH     Stretch final step this much to get to next vertex
                        (default: 1.0)
  --min-steps MIN_STEPS
                        How many step-length steps in a class are needed to
                        switch to that class - setting for all classes
                        (default: 1)
  --class-steps CLASS_STEPS
                        How many step-length steps in the last class defined
                        are needed to switch to that class - class specific
                        setting (default: [])
  --class CLASS_        Name of an output classification class (default: [])
  --values VALUES       Raster values (classes), space or comma separated, for
                        preceding output class. NoData for no data, * for all
                        remaining values (default: [])
  --fields FIELDS [FIELDS ...]
                        Field names, space or , separated, to copy from
                        `lines` input to output. (default: None)
  --range RANGE         Two raster values, space separated, min max, for
                        preceding output class NOT IMPLEMENTED (default: None)
  --band BAND           Raster band to use (default: 0)
  --get-classes         Just display a frequency table of classes seen
                        (default: False)
  --progress PROGRESS   Report lines proccessed every N lines (default: False)

Example

Outputs and output checksums from version 539cc34 Mon Apr 6 10:25:21 2015

python3 rasterlinefilter.py linestest/linestest.shp lctest/lctest.tif

testout --fields id lclass --get-classes

gives (sorted and reformatted):

21 164        43 49
22 20         52 163
23 28         71 33
24 5          81 94
31 43         90 529
41 537        95 121
42 50

then:

python3 rasterlinefilter.py linestest/linestest.shp lctest/lctest.tif \
    testout --fields id lclass \
    --class red --values 21,22,23,24,31,71 --class-steps 4 \
    --class blue --values 41,42,43,52,81,90,95 --class-steps 2

gives:

2 classes, 2 value lists, 2 class specific min-steps ok
'red', requires 4 steps in:
  [21, 22, 23, 24, 31, 71]
'blue', requires 2 steps in:
  [41, 42, 43, 52, 81, 90, 95]
75bcf5df869025cb211f6384eaa35d99ce79c842  testout/testout.dbf
f0cfd7b52fbd66c7f8ad7419bd5938137f3931e4  testout/testout.prj
ab6946d884f20c948a0287b186a9478619222280  testout/testout.shp
0c1d2d842b7ad48b61f2806166a151118bf31c1a  testout/testout.shx

Notes

• A line crossing a road or similar linear feature in a raster may or may not change class depending on the angle of intersection. Crossing at right angles will produce far fewer samples in the "road class" than crossing at an oblique angle.