zernike_moment.py

# import the necessary packages
from scipy.spatial import distance as dist
import numpy as np
import mahotas
import cv2
import imutils
 
def describe_shapes(image):
	# initialize the list of shape features
	shapeFeatures = []
 
	# convert the image to grayscale, blur it, and threshold it
	gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
	blurred = cv2.GaussianBlur(gray, (13, 13), 0)
	thresh = cv2.threshold(blurred, 50, 255, cv2.THRESH_BINARY)[1]
 
	# perform a series of dilations and erosions to close holes
	# in the shapes
	thresh = cv2.dilate(thresh, None, iterations=4)
	thresh = cv2.erode(thresh, None, iterations=2)
 
	# detect contours in the edge map
	cnts = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL,
		cv2.CHAIN_APPROX_SIMPLE)
	cnts = imutils.grab_contours(cnts)
 
	# loop over the contours
	for c in cnts:
		# create an empty mask for the contour and draw it
		mask = np.zeros(image.shape[:2], dtype="uint8")
		cv2.drawContours(mask, [c], -1, 255, -1)
 
		# extract the bounding box ROI from the mask
		(x, y, w, h) = cv2.boundingRect(c)
		roi = mask[y:y + h, x:x + w]
 
		# compute Zernike Moments for the ROI and update the list
		# of shape features
		features = mahotas.features.zernike_moments(roi, cv2.minEnclosingCircle(c)[1], degree=8)
		shapeFeatures.append(features)
 
	# return a tuple of the contours and shapes
	return (cnts, shapeFeatures)

# load the reference image containing the object we want to detect,
# then describe the game region
refImage = cv2.imread("zernike_reference.jpg")
(_, gameFeatures) = describe_shapes(refImage)
 
# load the shapes image, then describe each of the images in the image
shapesImage = cv2.imread("zernike_distractor.jpg")
(cnts, shapeFeatures) = describe_shapes(shapesImage)
 
# compute the Euclidean distances between the video game features
# and all other shapes in the second image, then find index of the
# smallest distance
D = dist.cdist(gameFeatures, shapeFeatures)
i = np.argmin(D)

# loop over the contours in the shapes image
for (j, c) in enumerate(cnts):
	# if the index of the current contour does not equal the index
	# contour of the contour with the smallest distance, then draw
	# it on the output image
	if i != j:
		box = cv2.minAreaRect(c)
		box = np.int0(cv2.cv.BoxPoints(box) if imutils.is_cv2() else cv2.boxPoints(box))
		cv2.drawContours(shapesImage, [box], -1, (0, 0, 255), 2)
 
# draw the bounding box around the detected shape
box = cv2.minAreaRect(cnts[i])
box = np.int0(cv2.cv.BoxPoints(box) if imutils.is_cv2() else cv2.boxPoints(box))
cv2.drawContours(shapesImage, [box], -1, (0, 255, 0), 2)
(x, y, w, h) = cv2.boundingRect(cnts[i])
cv2.putText(shapesImage, "FOUND!", (x, y - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.9,
	(0, 255, 0), 3)
 
# show the output images
cv2.imshow("Input Image", refImage)
cv2.imshow("Detected Shapes", shapesImage)
cv2.waitKey(0)