Counting features in satellite images using scikit-image

The example below uses scikit-image library to detect circular features in farms using center pivot irrigation in Saudi Arabia. It then counts and reports the number of farms. This is one of the ways in which libraries from the scientific Python ecosystem can be integrated with the ArcGIS platform.

It uses the Multispectral Landsat imagery available at ArcGIS Online.

Note: to run this sample, you need a few extra libraries in your conda environment. If you don't have the libraries, install them by running the following commands from cmd.exe or your shell

conda install scipy
conda install matplotlib
conda install scikit-image
from arcgis.gis import GIS
gis = GIS('', 'arcgis_python', 'amazing_arcgis_123')
l8 = gis.content.get('d9b466d6a9e647ce8d1dd5fe12eb434b')
l8lyr = l8.layers[0]
Multispectral Landsat
Landsat multispectral and multitemporal imagery with on-the-fly renderings and indices for visualization and analysis. The Landsat 8 imagery in this layer is updated daily and is directly sourced from the Landsat on AWS collection.Imagery Layer by esri_livingatlas
Last Modified: May 13, 2020
0 comments, 1 views

The code below sets the extent of the layer, to a known extent of farms in Saudi Arabia and then visualizes the landsat layer:

l8lyr.extent = {'spatialReference': {'latestWkid': 3857, 'wkid': 102100},
                                         'type': 'extent',
                                         'xmax': 4296559.143733407,
                                         'xmin': 4219969.241391764,
                                         'ymax': 3522726.823081019,
                                         'ymin': 3492152.0117669892}
<ImageryLayer url:"">

We can preprocess the imagery using raster functions. The code below uses the ndvi raster function to identify areas that have healthy vegetation. This preprocessing step makes the scikit-image blob detection algorithm work better.

from arcgis.raster.functions import ndvi, stretch
stretch(ndvi(l8lyr), stretch_type='PercentClip', min_percent=30, max_percent=70, dra=True)
<ImageryLayer url:"">

The code below exports the imagery to a file from which we read it using matplotlib's image API and plot it:

img = stretch(ndvi(l8lyr), stretch_type='PercentClip', min_percent=30, max_percent=70, dra=True).export_image(bbox=l8lyr.extent, bbox_sr=102100, size=[1200, 450],
                       export_format='jpeg', save_folder='.', save_file='centerpivotfarms.jpg', f='image')
import matplotlib.pyplot as plt
import matplotlib.image as mpimg

img = mpimg.imread('centerpivotfarms.jpg')

# what does it look like?
<matplotlib.figure.Figure at 0x1a44d2b8da0>

Blob detection using scikit-image

The code below uses scikit-image library to find blobs in the given grayscale image, and reports the number of farms thus detected. It also plots them for visualization using matplotlib.

Blobs are found using the Difference of Gaussian (DoG) method.

from skimage import feature, color
import matplotlib.pyplot as plt
bw = img.mean(axis=2)

fig = plt.figure(figsize = (15,15))
ax = fig.add_subplot(1,1,1)

blobs_dog = [(x[0],x[1],x[2]) for x in feature.blob_dog(-bw, 

#remove duplicates
blobs_dog = set(blobs_dog)

img_blobs = color.gray2rgb(img)

for blob in blobs_dog:
    y, x, r = blob
    c = plt.Circle((x, y), r+1, color='red', linewidth=2, fill=False)

plt.title('Center Pivot Farms')
print('Number of center pivot farms detected: ' + str(len(blobs_dog)))
<matplotlib.figure.Figure at 0x1a44d892e48>
Number of center pivot farms detected: 964

Your browser is no longer supported. Please upgrade your browser for the best experience. See our browser deprecation post for more details.