pixels-utils

A python wrapper to Sentera's public Titiler imagery API. The main purpose is to provide more direct access to the Pixels API via common data science libraries (like Rasterio and GeoPandas).

Setup and Installation (for development)

1. Set up SSH
2. Install pyenv and poetry.
3. Install package

git clone [email protected]:SenteraLLC/pixels-utils.git
cd pixels-utils
pyenv install $(cat .python-version)
poetry config virtualenvs.in-project true
poetry env use $(cat .python-version)
poetry install

4. Set up pre-commit to ensure all commits to adhere to black and PEP8 style conventions.

poetry run pre-commit install

Setup and Installation (used as a library)

If using pixels-utils as a dependency in your script, simply add it to the pyproject.toml in your project repo. Be sure to uuse the ssh: prefix so Travis has access to the repo for the library build process.

pyproject.toml

[tool.poetry.dependencies]
pixels_utils = { git = "ssh://[email protected]/SenteraLLC/pixels-utils.git", branch = "main", extras = ["rasterio"]}

Install pixels-utils and all its dependencies via poetry install.

poetry install

Catalog Documentation

Earth Search STAC API (maintained by Element 84)

Release/Tags

• A GitHub release is created on every push to the main branch using the create_github_release.yml Github Action Workflow
• Releases can be created manually through the GitHub Actions UI as well.
• The name of the Release/Tag will match the value of the version field specified in pyproject.toml
• Release Notes will be generated automatically and linked to the Release/Tag

Logging style

This library uses the "%s" logging.Formatter() style. For logging messages to show up, style should be set as style="%s" (this is the default). The recommended approach is to use the logging_init() function from py-utils.

For example:

import logging
from utils.logging.tqdm import logging_init

if __name__ == "__main__":
    logging_init(
        level=logging.INFO,
        format_string="{name} - {levelname}: {message}",
        style="%"
    )

Usage Example

Example 1 - Find all the scenes available for a geometry within a date range

scripts/scene_search.py

from pixels_utils.tests.data.load_data import sample_feature
from pixels_utils.scenes import parse_nested_stac_data, search_stac_scenes
from pixels_utils.stac_catalogs.earthsearch.v1 import EARTHSEARCH_URL, EarthSearchCollections


DATA_ID = 1

df_scenes = search_stac_scenes(
    geometry=sample_feature(DATA_ID),
    date_start="2019-01-01",
    date_end="2019-01-31",
    stac_catalog_url=EARTHSEARCH_URL,
    collection=EarthSearchCollections.sentinel_2_l2a,
    query={"eo:cloud_cover": {"lt": 80}},  # keeps less than 80% cloud cover,
    simplify_to_bbox=True,
)

print(
    df_scenes[["id"]]
    .merge(
        parse_nested_stac_data(df=df_scenes, column="properties")[["datetime", "eo:cloud_cover"]],
        left_index=True,
        right_index=True,
    )
    .to_markdown(tablefmt="pipe")
)

[OUTPUT]

	id	datetime	eo:cloud_cover
0	S2A_11TLM_20190110_0_L2A	2019-01-10T19:01:30.135000Z	26.9409
1	S2A_10TGS_20190110_0_L2A	2019-01-10T19:01:32.811000Z	61.8212
2	S2B_10TGS_20190125_0_L2A	2019-01-25T19:01:37.534000Z	55.6444

Example 2 - Get cloud-masked statistics for a geometry

scripts/statistics.py

from pixels_utils.stac_catalogs.earthsearch.v1 import expression_from_collection, EarthSearchCollections
from pixels_utils.tests.data.load_data import sample_feature, sample_scene_url
from pixels_utils.titiler import TITILER_ENDPOINT
from pixels_utils.titiler.endpoints.stac import QueryParamsStatistics, Statistics, StatisticsPreValidation
from pixels_utils.titiler.mask.enum_classes import Sentinel2_SCL_Group

DATA_ID = 1

scene_url = sample_scene_url(data_id=DATA_ID)

collection_ndvi = expression_from_collection(collection=EarthSearchCollections.sentinel_2_l2a, spectral_index="NDVI")

query_params = QueryParamsStatistics(
    url=scene_url,
    feature=sample_feature(DATA_ID),
    assets=None,  # ["nir"]
    expression=collection_ndvi.expression,  # "(nir-red)/(nir+red)"
    asset_as_band=True,
    asset_bidx=None,
    coord_crs=None,
    max_size=None,
    height=None,
    width=None,
    gsd=None,
    nodata=None,
    unscale=None,
    resampling=None,
    categorical=None,
    c=None,
    p=None,
    histogram_bins=None,
    histogram_range=None,
)

# Raises an AssertionError if any of the assets are not available for the query_params
# If you get a message "StatisticsPreValidation passed: all required assets are available.", you can proceed to Statistics
stats_preval = StatisticsPreValidation(query_params, titiler_endpoint=TITILER_ENDPOINT)


# Now actually request Statistics - for only arable pixels (whitelist=True)!
stats_arable_wlist = Statistics(
    query_params=query_params,  # collection_ndvi.expression - "(nir-red)/(nir+red)"
    titiler_endpoint=TITILER_ENDPOINT,
    mask_enum=Sentinel2_SCL_Group.ARABLE,
    mask_asset="scl",
    whitelist=True,
)

stats_arable_wlist.response.json()

[OUTPUT]

root - INFO - Item "S2B_10TGS_20220419_0_L2A" asset is AVAILABLE: "red".
root - INFO - Item "S2B_10TGS_20220419_0_L2A" asset is AVAILABLE: "nir".
root - INFO - StatisticsPreValidation PASSED. All required assets are available.
{'type': 'Feature',
 'geometry': {'type': 'MultiPolygon',
  'coordinates': [[[[-119.036182, 46.239917],
     [-119.044517, 46.237081],
     [-119.044048, 46.239172],
     [-119.0413, 46.240763],
     [-119.036182, 46.239917]]]]},
 'properties': {'statistics': {'where(scl==4,(nir-red)/(nir+red),where(scl==5,(nir-red)/(nir+red),0.0))': {'min': 0.105562855891371,
    'max': 0.24060150375939848,
    'mean': 0.14380639335058598,
    'count': 1021.0,
    'sum': 146.82632761094828,
    'std': 0.01212996991547649,
    'median': 0.14462416745956233,
    'majority': 0.14285714285714285,
    'minority': 0.105562855891371,
    'unique': 801.0,
    'histogram': [<omitted for brevity>],
    'valid_percent': 40.71,
    'masked_pixels': 1487.0,
    'valid_pixels': 1021.0,
    'percentile_98': 0.16414680642034107,
    'percentile_2': 0.11589278323054499}}}}