canny edge detection and gaussian blur

edge_detection.py

# Running first a gaussian blur (kernel size = 3) 
# and then Canny edge detection (low/high = 1:2 or 1:3) 
# is a handy way to get the boundaries
from matplotlib import image
from matplotlib import pyplot
import numpy
import cv2

view_from_windshield = image.imread('test.jpg')
working_copy = numpy.copy(view_from_windshield)

# first convert to grayscale
grayscale = cv2.cvtColor(working_copy, cv2.COLOR_RGB2GRAY)
#apply gaussian blur before canny edge detection
kernel_size=3
gaussian_blurred = cv2.GaussianBlur(grayscale,(kernel_size, kernel_size), 0)
# now apply canny edge detection threshold ratio 1:3
low_threshold = 60
high_threshold = 180
edges = cv2.Canny(gaussian_blurred, low_threshold, high_threshold)

#pyplot.imshow(grayscale, cmap='gray')
#pyplot.imshow(gaussian_blurred, cmap='gray')
pyplot.imshow(edges, cmap='Greys_r')
pyplot.show()
pyplot.imsave('edges.jpg', edges, cmap='Greys_r')

After we have marked the boundaries in an image, we can identify various shapes. Hough Transform is one of such trick to identify the shapes, which we shall explore next.