def get_centerline(B3, B6, size):
"""
Using RivaMap library finds the largest centerline from
Bands 3 and 6 lansat 8 imagery
"""
I1 = preprocess.mndwi(B3, B6)
filters = singularity_index.SingularityIndexFilters()
psi, widthMap, orient = singularity_index.applyMMSI(I1, filters)
nms = delineate.extractCenterlines(orient, psi)
return delineate.getBiggestCenterline(nms, size)
def test_applyMMSI_scales():
'''
Testing with more than just 1 scale size
'''
t_input = np.array([[20,20,50,20,20],[20,20,55,20,20],[20,20,50,20,20]], dtype='uint8')
filts = singularity_index.SingularityIndexFilters(nrScales=2)
[psi, widthMap, orient] = singularity_index.applyMMSI(t_input,filts)
# expect max response in the center
psi_max = np.max(psi)
psi_bool = psi==psi_max
# make assertion
assert psi_bool[1,2] == True
def batch_compute(landsat_images):
print len(landsat_images)
filters = singularity_index.SingularityIndexFilters(minScale=1.2, nrScales=14)
for landsat_image in landsat_images:
temp_dir = os.path.join(save_dir, 'temp')
if not os.path.exists(temp_dir):
os.makedirs(temp_dir)
with zipfile.ZipFile(landsat_image) as zf:
for zippedname in zf.namelist():
if zippedname.endswith('.tif'):
zf.extract(zippedname, path = temp_dir)
temp_im_dir = os.path.join(temp_dir, zippedname)
I1 = cv2.imread(temp_im_dir, cv2.IMREAD_UNCHANGED)
gm = georef.loadGeoMetadata(temp_im_dir)
#delete the temp file
os.remove(temp_im_dir)
psi, widthMap, orient = singularity_index.applyMMSI(I1, filters)
nms = delineate.extractCenterlines(orient, psi)
centerlines = delineate.thresholdCenterlines(nms, tLow=0.012, tHigh=0.11)
# remove the overlapping response
numRow, numCol = centerlines.shape
padRow = numRow/10
padCol = numCol/10
centerlines[0:padRow, :] = 0
centerlines[:, 0:padCol] = 0
centerlines[-padRow:, :] = 0
centerlines[:, -padCol:] = 0
#save results
wrsname = ntpath.basename(landsat_image)
georef.exportCSVfile(orient, psi, centerlines, widthMap, gm, os.path.join(save_dir, wrsname[:-4] + '.csv'))
#You can download the example images from AWS:
#http://landsat-pds.s3.amazonaws.com/L8/138/045/LC81380452015067LGN00/LC81380452015067LGN00_B3.TIF
#http://landsat-pds.s3.amazonaws.com/L8/138/045/LC81380452015067LGN00/LC81380452015067LGN00_B6.TIF
# Read bands 3 and 6 of an example Landsat 8 image
B3 = cv2.imread("LC81380452015067LGN00_B3.TIF", cv2.IMREAD_UNCHANGED)
B6 = cv2.imread("LC81380452015067LGN00_B6.TIF", cv2.IMREAD_UNCHANGED)
# Compute the modified normalized difference water index of the input
I1 = preprocess.mndwi(B3, B6)
# Create the filters that are needed to compute the singularity index
filters = singularity_index.SingularityIndexFilters()
# Compute the modified multiscale singularity index
psi, widthMap, orient = singularity_index.applyMMSI(I1, filters)
# Extract channel centerlines
nms = delineate.extractCenterlines(orient, psi)
centerlines = delineate.thresholdCenterlines(nms)
# Generate a raster map of the extracted channels
raster = visualization.generateRasterMap(centerlines, orient, widthMap)
# Generate a vector map of the extracted channels
visualization.generateVectorMap(centerlines,
orient,
widthMap,
saveDest="vector.pdf")
# Generate a quiver plot
#prints names of files being used
for infile in B3s:
print infile
for infile in B6s:
print infile
# channel response change
for i in range(0, len(B3s)):
print(i)
filters = singularity_index.SingularityIndexFilters()
B3 = cv2.imread(B3s[i], cv2.IMREAD_UNCHANGED)
B6 = cv2.imread(B6s[i], cv2.IMREAD_UNCHANGED)
I1 = preprocess.mndwi(B3, B6)
psi, widthMap, orient = singularity_index.applyMMSI(I1,
filters,
togglePolarity=False)
nms = delineate.extractCenterlines(orient, psi)
centerlines = delineate.thresholdCenterlines(nms)
print("creating files")
gm = georef.loadGeoMetadata(B3s[i])
base = os.path.basename(B3s[i])
georef.exportCSVfile(centerlines, widthMap, gm,
str(base) + "_geo_points.csv")
psi = preprocess.contrastStretch(psi)
psi = preprocess.double2im(psi, 'uint16')
georef.saveAsGeoTiff(gm, psi, str(base) + "_geo_psi.TIF")
mndwi = preprocess.contrastStretch(I1)
#You can download the example images from AWS:
#http://landsat-pds.s3.amazonaws.com/L8/138/045/LC81380452015067LGN00/LC81380452015067LGN00_B3.TIF
#http://landsat-pds.s3.amazonaws.com/L8/138/045/LC81380452015067LGN00/LC81380452015067LGN00_B6.TIF
# Read bands 3 and 6 of an example Landsat 8 image
B3 = cv2.imread("LC81380452015067LGN00_B3.TIF", cv2.IMREAD_UNCHANGED)
B6 = cv2.imread("LC81380452015067LGN00_B6.TIF", cv2.IMREAD_UNCHANGED)
# Compute the modified normalized difference water index of the input
I1 = preprocess.mndwi(B3, B6)
# Create the filters that are needed to compute the singularity index
filters = singularity_index.SingularityIndexFilters()
# Compute the modified multiscale singularity index
psi, widthMap, orient = singularity_index.applyMMSI(I1, filters)
# Extract channel centerlines
nms = delineate.extractCenterlines(orient, psi)
centerlines = delineate.thresholdCenterlines(nms)
# Generate a raster map of the extracted channels
raster = visualization.generateRasterMap(centerlines, orient, widthMap)
# Generate a vector map of the extracted channels
visualization.generateVectorMap(centerlines, orient, widthMap, saveDest = "vector.pdf")
# Generate a quiver plot
visualization.quiverPlot(psi, orient, saveDest = "quiver.pdf")
# Save the images that are created at the intermediate steps
