redlight_detector

This is an OpenCV project which aims to filter the red color in a given png file.

Tutorial

Code

Firstly we need to import the necessary packages. Importing numpy for numerical processing, argparse for parsing our command line arguments and cv2 for OpenCV operations.

import numpy as np
import cv2, argparse

Lines below are handling the command line arguments, where we will be fetching our png file.

ap = argparse.ArgumentParser()
ap.add_argument("-i", "--image")
args = vars(ap.parse_args())

img = cv2.imread(args["image"])

Now we have to define the thresholds (or boundaries) for the red color. Since OpenCV represents images as NumPy arrays in reverse order, instead of RGB, we can think of our array as BGR (Blue, Green, Red).

Blue and Green colors must be between [0,25], since we don't want their effect on image too much. Red color may be any number since we want to get all the tones of red. ([0,255])

lower = np.array([0, 0, 0], dtype = "uint8")
upper = np.array([25, 25, 255], dtype = "uint8")

And we implement our mask by using these boundaries, then output a resulting image by filtering (bitwise AND) operation.

mask = cv2.inRange(img, lower, upper)

output = cv2.bitwise_and(img, img, mask=mask)

Now we can show our output beside the original image.

cv2.imshow("", np.hstack([img, output]))

How to run it?

We have to execute the following command line to run our program where redlight_detector.py is the python script and img.png is the image to be filtered by the program.

python redlight_detector.py --image img.png

Executing the command line above with an image -named img.png- containing a red light will produce an output like this: