def __init__(self, name, path, k): self.name = name # name of image self.path = path # path of image self.array,_ = OrthoImage.load(path) # numpy array of raster self.label = None # clustering of image self.cover = np.zeros(k) # cover rates of cluster self.error = np.zeros(k) # error rates of cluster
def readPolygonTIF(polygonTIF, shpfile, down): """ Returns the x, y coordinates of the pixels as well as the class labels in numpy arrays based off the polygonTIF. polygonTIF will be effectively downsampled in this function, and thus so are x, y, and class. """ sample = 2**down # arrays to be used to find coordinates and type of the pixels x = np.zeros(0) y = np.zeros(0) classes = np.zeros(0) # read in the polygonImage polyImage,_ = OrthoImage.load(polygonTIF) polyImage = GDAL2OpenCV(polyImage) # downsample the polTIF if down > 0: polyImage = pyramid(polyImage, down) # dimensions of TIF image if len(polyImage.shape) == 3: h, w,_ = polyImage.shape elif len(polyImage.shape) == 2: h, w = polyImage.shape # downsampled dimensions of the bounding box of the polygons sf = shapefile.Reader(shpfile) lx,_ ,_ , uy = sf.bbox lx = lx/sample uy = uy/sample*-1 # iterate through the entire polyImage for i in xrange(h): for j in xrange(w): # if the value of the pixel is 255(crop) then add to the arrays # as a crop pixel. The condition is adjusting for downsampling. if polyImage[i, j] > 1: x = np.append(x, j + lx) y = np.append(y, i + uy) classes = np.append(classes, 1) # else if the value of the pixel is 1 (noncrop) then add to arrays # as a noncrop pixel elif polyImage[i, j] == 1: x = np.append(x, j + lx) y = np.append(y, i + uy) classes = np.append(classes, 2) return x, y, classes
