ee.Image.sample

  • The sample method samples pixels from an image and returns them as a FeatureCollection, with each feature having one property per band.

  • By default, the sample method drops features that intersect masked pixels, which result in null-valued properties.

  • The sample method has various arguments to control the sampling process, including the region, scale, projection, subsampling factor or number of pixels, randomization seed, whether to drop nulls, tile scale, and whether to include geometries.

  • The sample method returns a FeatureCollection.

Samples the pixels of an image, returning them as a FeatureCollection. Each feature will have 1 property per band in the input image. Note that the default behavior is to drop features that intersect masked pixels, which result in null-valued properties (see dropNulls argument).

UsageReturns
Image.sample(region, scale, projection, factor, numPixels, seed, dropNulls, tileScale, geometries)FeatureCollection
ArgumentTypeDetails
this: imageImageThe image to sample.
regionGeometry, default: nullThe region to sample from. If unspecified, uses the image's whole footprint.
scaleFloat, default: nullA nominal scale in meters of the projection to sample in.
projectionProjection, default: nullThe projection in which to sample. If unspecified, the projection of the image's first band is used. If specified in addition to scale, rescaled to the specified scale.
factorFloat, default: nullA subsampling factor, within (0, 1]. If specified, 'numPixels' must not be specified. Defaults to no subsampling.
numPixelsLong, default: nullThe approximate number of pixels to sample. If specified, 'factor' must not be specified.
seedInteger, default: 0A randomization seed to use for subsampling.
dropNullsBoolean, default: truePost filter the result to drop features that have null-valued properties.
tileScaleFloat, default: 1A scaling factor used to reduce aggregation tile size; using a larger tileScale (e.g., 2 or 4) may enable computations that run out of memory with the default.
geometriesBoolean, default: falseIf true, adds the center of the sampled pixel as the geometry property of the output feature. Otherwise, geometries will be omitted (saving memory).

Examples

Code Editor (JavaScript)

// Demonstrate extracting pixels from an image as features with
// ee.Image.sample(), and show how the features are aligned with the pixels.

// An image with one band of elevation data.
var image = ee.Image('CGIAR/SRTM90_V4');
var VIS_MIN = 1620;
var VIS_MAX = 1650;
Map.addLayer(image, {min: VIS_MIN, max: VIS_MAX}, 'SRTM');

// Region to sample.
var region = ee.Geometry.Polygon(
  [[[-110.006, 40.002],
    [-110.006, 39.999],
    [-109.995, 39.999],
    [-109.995, 40.002]]], null, false);
// Show region on the map.
Map.setCenter(-110, 40, 16);
Map.addLayer(ee.FeatureCollection([region]).style({"color": "00FF0022"}));

// Perform sampling; convert image pixels to features.
var samples = image.sample({
  region: region,

  // Default (false) is no geometries in the output.
  // When set to true, each feature has a Point geometry at the center of the
  // image pixel.
  geometries: true,

  // The scale is not specified, so the resolution of the image will be used,
  // and there is a feature for every pixel. If we give a scale parameter, the
  // image will be resampled and there will be more or fewer features.
  //
  // scale: 200,
});

// Visualize sample data using ee.FeatureCollection.style().
var styled = samples
  .map(function (feature) {
    return feature.set('style', {
      pointSize: feature.getNumber('elevation').unitScale(VIS_MIN, VIS_MAX)
          .multiply(15),
    });
  })
  .style({
    color: '000000FF',
    fillColor: '00000000',
    styleProperty: 'style',
    neighborhood: 6,  // increase to correctly draw large points
  });
Map.addLayer(styled);

// Each sample feature has a point geometry and a property named 'elevation'
// corresponding to the band named 'elevation' of the image. If there are
// multiple bands they will become multiple properties. This will print:
//
// geometry: Point (-110.01, 40.00)
// properties:
//   elevation: 1639
print(samples.first());

Python setup

See the Python Environment page for information on the Python API and using geemap for interactive development.

import ee
import geemap.core as geemap

Colab (Python)

# Demonstrate extracting pixels from an image as features with
# ee.Image.sample(), and show how the features are aligned with the pixels.

# An image with one band of elevation data.
image = ee.Image('CGIAR/SRTM90_V4')
vis_min = 1620
vis_max = 1650
m = geemap.Map()
m.add_layer(image, {'min': vis_min, 'max': vis_max}, 'SRTM')

# Region to sample.
region = ee.Geometry.Polygon(
    [[
        [-110.006, 40.002],
        [-110.006, 39.999],
        [-109.995, 39.999],
        [-109.995, 40.002],
    ]],
    None,
    False,
)
# Show region on the map.
m.set_center(-110, 40, 16)

m.add_layer(ee.FeatureCollection([region]).style(color='00FF0022'))

# Perform sampling convert image pixels to features.
samples = image.sample(
    region=region,
    # Default (False) is no geometries in the output.
    # When set to True, each feature has a Point geometry at the center of the
    # image pixel.
    geometries=True,
    # The scale is not specified, so the resolution of the image will be used,
    # and there is a feature for every pixel. If we give a scale parameter, the
    # image will be resampled and there will be more or fewer features.
    #
    # scale=200,
)


def scale_point_size(feature):
  elevation = feature.getNumber('elevation')
  point_size = elevation.unitScale(vis_min, vis_max).multiply(15)
  feature.set('style', {'pointSize': point_size})
  return feature


# Visualize sample data using ee.FeatureCollection.style().
styled = samples.map(scale_point_size).style(
    color='000000FF',
    fillColor='00000000',
    styleProperty='style',
    neighborhood=6,  # increase to correctly draw large points
)
m.add_layer(styled)
display(m)

# Each sample feature has a point geometry and a property named 'elevation'
# corresponding to the band named 'elevation' of the image. If there are
# multiple bands they will become multiple properties. This will print:
#
# geometry: Point (-110.01, 40.00)
# properties:
#   elevation: 1639
display(samples.first())