def get_incorrect_pixels(inputpath, referencepath, vitimeseriespath, startdoy, interval, nodata=None):
#Open images
inputimage = openImage(inputpath)
referenceimage = openImage(referencepath)
timeseries = openImage(vitimeseriespath)
#Get band 1 of input and reference images and read as arrays
inband = inputimage[5].GetRasterBand(1)
inarray = inband.ReadAsArray(0, 0, inputimage[0], inputimage[1])
inband = ""
refband = referenceimage[5].GetRasterBand(1)
refarray = refband.ReadAsArray(0, 0, referenceimage[0], referenceimage[1])
refband = ""
#Close in and ref images
#Find unique values in the input image
uniquevals = set(numpy.unique(inarray).tolist())
uniquevals.remove(nodata)
print uniquevals
#Iterate through pixels in images, comparing arrays for each value in input image, making note of tested values and excluding them from comparison with 0 value in input image
#Return pixel coordinates and the identified and true identities of the pixels from the input and reference images
coordinatelist = {}
for value in uniquevals:
coordinatelist[value] = []
numincorrect = 0
for col in range(0, inputimage[0]):
for row in range(0, inputimage[1]):
if (inarray[row, col] != 0) and (inarray[row, col] != refarray[row, col]) and (inarray[row, col] != nodata):
if refarray[row, col] in uniquevals:
coordinatelist[refarray[row, col]].append([col, row, inarray[row, col], refarray[row, col]])
numincorrect += 1
else:
coordinatelist[0].append([col, row, inarray[row, col], refarray[row, col]])
numincorrect += 1
elif (inarray[row,col] == 0) and (refarray[row, col] in uniquevals):
coordinatelist[refarray[row, col]].append([col, row, inarray[row, col], refarray[row, col]])
numincorrect += 1
#For each returned pixel, extract and plot the curve from vitimeseries and export plot as image; be sure to label plot correctly
bandcnt = timeseries[5].RasterCount
arrays = {}
for i in range(1, bandcnt + 1):
band = timeseries[5].GetRasterBand(i)
data = band.ReadAsArray(0, 0, timeseries[0], timeseries[1])
arrays[i] = data
print("Found {0} incorrect pixels to process...".format(numincorrect))
def plot_points(multidateraster, pointfile, startdoy, doyinterval):
"""
"""
import os
from utils import unique_name
from plotting import PixelPlot
from core import pixel as pixelObject
from vectorFunctions import get_px_coords_from_shapefile
from imageFunctions import openImage
outpath = unique_name(os.path.dirname(multidateraster), "plots", ext=".pdf", usetime=True)
coords = get_px_coords_from_shapefile(multidateraster, pointfile)
plot = PixelPlot(os.path.dirname(outpath), os.path.basename(outpath))
raster = openImage(multidateraster)
for coord in coords:
pixel = pixelObject(coord[0], coord[1])
pixel.get_pixel_values(raster, startdoy, doyinterval)
plot.add_pixel(pixel, closefigure=True)
plot.close_plot()
raster = None
def get_px_coords_from_shapefile(raster, shapefile):
"""
Takes geographic coordinates from a shapefile and finds the corresponding pixel coordinates on a raster.
rst = "/Users/phoetrymaster/Documents/School/Geography/Thesis/Data/MODIS_KANSAS_2007-2012/reprojected/clips/KansasEVI_2012_clip1.tif"
#rst = "/Users/phoetrymaster/Documents/School/Geography/Thesis/Data/polygonclip_20130929223024_325071991/resampled/newclips/2012clip1.tif"
shp = "/Users/phoetrymaster/Documents/School/Geography/Thesis/Data/MODIS_KANSAS_2007-2012/SampleAreas/samplepoints2012_clip1_new.shp"
print get_px_coords_from_shapefile(rst, shp)
"""
#TODO docstrings
from imageFunctions import openImage
# open, close image file and get properties
raster = openImage(raster)
imageproperties = gdalProperties(raster)
rasterwkt = raster.GetProjectionRef()
oSRSop = osr.SpatialReference()
oSRSop.ImportFromWkt(rasterwkt)
raster = None
shppoints = load_points(shapefile, oSRSop)
# get pixel coords from point coords
pxcoords = get_px_coords_from_geographic_coords(imageproperties, shppoints)
return pxcoords
def fit_refs_to_image(imagetoprocess, outputdirectory, signaturecollection, startDOY,
doyinterval, bestguess, threshold=None, ndvalue=-3000, fitmethod=None, subset=None,
meantype=None, workers=4, timebounds=None, xbounds=None, ybounds=None):
"""
imagepath = "/Users/phoetrymaster/Documents/School/Geography/Thesis/Data/ARC_Testing/ClipTesting/ENVI_1/test_clip_envi_3.dat"
outdir = "/Users/phoetrymaster/Documents/School/Geography/Thesis/Data/OutImages/"
newfoldername = "Testing"
drivercode1 = 'ENVI'
#ndvalue = -3000
startDOY = 1
interval = 16
threshold = 500
bestguess = 0
fitmthd = 'SLSQP'
mean = geometric # Acceptable values are geometric (geometric mean) and arithmetic (arithmetic mean). It is an optional argument for the classifier.
refs = {
'soy': {1: 174.5, 97: 1252.25, 65: 1139.5, 209: 7659.0, 273: 4606.75, 337: 1371.75, 17: 1055.5, 33: 1098.0,
49: 1355.25,
129: 1784.75, 257: 6418.0, 321: 1644.5, 305: 1472.75, 193: 5119.75, 289: 1878.75, 177: 3439.5, 241: 7565.75,
81: 1205.5, 225: 7729.75, 145: 1736.25, 161: 1708.25, 353: 1358.25, 113: 1340.0},
'corn': {1: 392.25, 97: 1433.25, 65: 1258.5, 209: 6530.0, 273: 1982.5, 337: 1658.5, 17: 1179.25, 33: 1196.75,
49: 1441.25, 129: 1885.25, 257: 2490.25, 321: 1665.75, 305: 1439.0, 193: 6728.25, 289: 1634.5,
177: 6356.75,
241: 4827.25, 81: 1355.75, 225: 5547.5, 145: 2196.5, 161: 3143.25, 353: 1704.75, 113: 1716.5},
'wheat': {1: 719.75, 97: 6594.75, 65: 1935.25, 209: 2013.5, 273: 1493.5, 337: 1498.25, 17: 1816.5, 33: 1815.0,
49: 1985.25, 129: 6758.0, 257: 1685.75, 321: 1582.5, 305: 1163.25, 193: 2186.25, 289: 1264.5, 177: 2222.5,
241: 2301.0, 81: 4070.5, 225: 1858.0, 145: 6228.5, 161: 3296.5, 353: 1372.5, 113: 7035.25}
}
sys.exit(fit_refs_to_image(imagepath, outdir, newfoldername, refs, startDOY, interval, threshold, bestguess, fitmthd, meantype=mean))
"""
#TODO docstrings
start = dt.now()
print(start)
try:
print("\nProcessing {0}...".format(imagetoprocess))
print("Outputting files to {0}\n".format(outputdirectory))
#Open multi-date image to analyze
image = openImage(imagetoprocess)
imageproperties = gdalProperties(image)
print("Input image dimensions are {0} columns by {1} rows and contains {2} bands.".format(imageproperties.cols,
imageproperties.rows,
imageproperties.bands))
array = read_image_into_array(image) # Read all bands into a 3d array representing the image stack (x, y, time orientation)
image = ""
if timebounds:
timebounds = (bestguess + timebounds[0], bestguess + timebounds[1])
else:
timebounds = (bestguess - 10, bestguess + 10)
if not xbounds:
xbounds = (0.6, 1.4)
if not ybounds:
ybounds = (0.6, 1.4)
bounds = (xbounds, ybounds, timebounds)
print(bounds)
if subset:
subset = get_px_coords_from_shapefile(imagetoprocess, subset)
processes = []
for signum, signature in enumerate(signaturecollection.signatures, start=1):
p = multiprocessing.Process(target=process_reference,
args=(outputdirectory, signature, array, imageproperties, startDOY, doyinterval,
bestguess, ndvalue),
kwargs={"subset": subset, "fitmthd": fitmethod, "meantype": meantype,
"thresh": threshold, "bounds": bounds})
#TODO: Problem with joining/starting processes--original thread closes before others are completed -- believe this is now fixed.
p.start()
processes.append(p)
if len(processes) == workers:
for p in processes:
p.join()
processes.remove(p)
for p in processes:
p.join()
except Exception as e:
import traceback
exc_type, exc_value, exc_traceback = sys.exc_info()
print(e)
traceback.print_exception(exc_type, exc_value, exc_traceback, limit=2, file=sys.stdout)
finally:
print(dt.now() - start)
def classify_and_assess_accuracy(outputdir, cropimgpath, searchstringsvals, filevalist, nodata, thresholdsandlength,
classifiedimagename=None, numberofprocesses=4):
"""
"""
#TODO Docstring
#from multiprocessing import Pool
#pool = Pool(numberofprocesses)
if classifiedimagename is None:
today = dt.now()
classifiedimagename = today.strftime("%Y-%m-%d_%H%M_") + os.path.splitext(os.path.basename(cropimgpath))[0]
classificationimage = os.path.join(outputdir, classifiedimagename + ".tif")
accuracyimage = os.path.join(outputdir, classifiedimagename + "_accuracy.tif")
accuracyreport = os.path.join(outputdir, classifiedimagename + ".txt")
#np.set_printoptions(threshold=np.nan) # For debug: Makes numpy print whole contents of an array.
#Crop image is constant for all iterations
cropimg = openImage(cropimgpath)
cropimgproperties = gdalProperties(cropimg)
croparray = read_image_into_array(cropimg)
cropimg = None
arraylist = [(read_image_into_array(openImage(f[0])), f[1]) for f in filevalist] # fit images with crop vals
writestring = "The fit rasters and truth values used for this classification process are:\n"
for f in filevalist:
writestring = writestring + "\t{0}\t{1}\n".format(f[0], f[1])
writestring = writestring + "\n"
bestacc = 0
bestthresh = None
thresholdlist, lengthofthresholdlist = thresholdsandlength
try:
if lengthofthresholdlist == 1:
writestring = writestring + "\n\n**Only using a single threshold value--not iterating.**\n\n"
bestthresh = thresholdlist[0]
else:
#TODO: Refactor to allow use of multiprocessing.Pool.map -- need to reason about the output/logging
for thresh in thresholdlist:
start = dt.now()
accuracy, classification, outstring = classify_with_threshold(croparray, arraylist, searchstringsvals,
thresh, nodata, cropimgproperties.nodata)
writestring = writestring + outstring
if accuracy > bestacc:
bestacc = accuracy
bestthresh = thresh
elapsed = dt.now() - start
toprint = [thresh, "{}:{}".format(elapsed.seconds, str(elapsed.microseconds).zfill(6)), accuracy, bestacc, bestthresh]
width = (6 * len(arraylist))
sys.stdout.write("Thresh: {: <{width}} Time: {} Acc: {: <14} Best: {: <14} at {}\r".format(*toprint, width=width))
sys.stdout.flush()
except Exception as e:
import traceback
exc_type, exc_value, exc_traceback = sys.exc_info()
print e
traceback.print_exception(exc_type, exc_value, exc_traceback, limit=2, file=sys.stdout)
finally:
accuracy, classificationarray, outstring = classify_with_threshold(croparray, arraylist, searchstringsvals,
bestthresh, nodata, cropimgproperties.nodata)
writestring = writestring + outstring
accuracyarray = find_correct_incorrect_array(croparray, classificationarray, ndvalue=nodata,
truthndvalue=cropimgproperties.nodata)
with open(accuracyreport, 'w') as text:
text.write("Classification using fit images from {0}\n\n".format(os.path.dirname(filevalist[0][0])))
text.write("{0}\nBest:\n{1} {2}".format(writestring, bestthresh, accuracy))
print("\n{0}, {1}".format(bestthresh, accuracy))
driver = gdal.GetDriverByName("ENVI")
driver.Register()
write_output_image(cropimgproperties, classificationimage, classificationarray, nodata)
write_output_image(cropimgproperties, accuracyimage, accuracyarray, nodata)
print("outputted")
return 0
