brendonhall/fieldfinder

FieldFinder

Fieldfinder is a Python module for detecting agricultural development in 8-band PlanetScope images. Information about land cover can be extracted from the different wavelength bands in a multispectral image. Spectral Indices are combinations and ratios of the different bands, and can be used as features to identify land coverage. A common index for identifying vegetation is the normalized difference vegetation index (NDVI). The measure scales with the presence of live, green vegetation. The formula for NDVI is: $$NDVI = \frac{NIR - R}{NIR + R}$$ NDVI is the ratio of the difference of the red (R) and near-infrared (NIR) radiances over their sum as a means to adjust for or “normalize” the effects of the solar zenith angle. Live green plants appear relatively dark in the photosynthetically active radiation (PAR) band and relatively bright in the near-infrared. Thus high values of NDVI would indicate dense vegetation (ie: agriculture).

Note: There are many spectral indices besides NDVI (Agapiou 2012), and fieldfinder can be easily expanded to accomodate these.

Radiance vs. Reflectance

Planet Labs' analytic data products are reported in units of radiance: $W*m^{-2}*sr^{-1}$. This quantifies the amount of light captured over a given spot on the ground. This depends on the amount and frequencies of light that are reflected from the surface, but also on satellite altitude, time of day, time of year and individual satellite characteristics. To compare spectral indices across many images, radiance values should be converted to Top of Atmosphere (TOA) reflectance to provide an apples-to-apples comparison

Installation

The easiest way to install fieldfinder is using pip:

pip install fieldfinder

You can also install fieldfinder from source. Clone the fieldfinder repo. Go into the fieldfinder directory and run:

pip install -e .

Quickstart

fieldfinder is designed to calculate a spectral index (such as NDVI) from an 8-band PlanetScope AnalyticMS image, and output a raster mask that indicates where the mask exceeds a given threshold. This can be used to indicate regions with heavy vegetation, such as agricultural fields. The following example demonstrates how to create an NVDI spectral index from an 8-band PlanetScope image, and output a mask file with values of 255 where NDVI > 0.65, and zero otherwise. The output file is reprojected to lat/long (EPSG:4326) coordinates.

from fieldfinder import SpectralIndex

SpectralIndex.create_mask_file(
    filename = 'example_AnalyticMS_8b.tif',
    output_file = 'example_NDVI_mask.tif',
    threshold=0.65,
    out_proj = 'EPSG:4326',
    index_type = 'ndvi'
)

For a more detailed tutorial on using fieldfinder to calculate a NDVI indicator for agricultural land use based on an 8-band PlanetScope image, see the tutorial notebook.