Is the battery charged? Is the broadband connection up or down?
If you have red/green color blindness, these questions can sometimes be hard to answer.
This is the repository for a simple mobile web app that analyzes video from your phone's camera and labels regions of bright red or green (or blue) color. Point the camera at a red or green (or blue) LED and it should be labeled.
Works best with good lighting. You may need to get a close-up view and/or try different angles before the color is detected. Minimal effort is made to avoid labeling things that aren't LEDs.
Try it here https://www.isitredorgreen.com
General color finding apps are available but I have found them lacking for reading LEDs:
- they only sample color from a small target area and are hard to aim; making this worse, LEDs tend to be overexposed by the camera and must be sampled around the perimeter to get an accurate color reading
- they don't allow you to read multiple LEDs at the same time
- they don't handle blinking LEDs well
- they display colors from a large palette of color names instead of giving a simple red/green answer
What's needed, I thought, is a simple app that does just one thing: point the camera at some LEDs and the app finds and labels them red or green (or blue). Surely this must be possible with today's computer vision technology.
So, I grabbed OpenCV, spent some quality time with google and stackoverflow, tried several approaches, and settled on a simple algorithm that seems to work pretty well. I'm publishing the code here, both the python that I developed the algorithm in and an html5 javascript port that runs in mobile web browsers.
The algorithm is simple:
- look for all pixels with a red component that is both bright (200-255) and at least twice as large as the green or blue components
- find contour lines around those pixels, grouping them into regions
- label those regions with a red R
- do the same for green and blue
In python, using opencv-python v4.5, the algorithm looks like this
def labelColor(src, dst, lowerBound, upperBound, text, textColor):
# create a mask including all pixels within the specified color range
mask = cv2.inRange(src, lowerBound, upperBound)
# find contours around clusters of pixels selected in mask
contours, hierarchy = cv2.findContours(mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
# label each contour with the specified text
for c in contours:
M = cv2.moments(c)
if M['m00'] != 0:
# place label at the centroid of the contour
cx = int(M['m10']/M['m00'])
cy = int(M['m01']/M['m00'])
cv2.putText(dst, text, (cx, cy), cv2.FONT_HERSHEY_SIMPLEX, 2, textColor, 4)
# colors are BGR
labelColor(frame, dst, (0, 0, 200), (100, 100, 255), 'R', (0, 0, 255));
labelColor(frame, dst, (0, 200, 0), (100, 255, 100), 'G', (0, 255, 0));
labelColor(frame, dst, (200, 0, 0), (255, 100, 100), 'B', (255, 0, 0));
A command-line python program runs the algorithm on photos and videos, displaying the labeled output in a window (unless the -b option is specified) and also saving it to a file.
% ./labelRGB.py -h
Usage: labelRGB.py [options] file
Options:
-h, --help show this help message and exit
-b non-interactive batch mode
-w wait for key press before exit
Label photo
% ./labelRGB.py charger.jpg
Label video
% ./labelRGB.py charger.avi
An html5 javascript port, using opencv.js v4.5, runs the algorithm in a mobile web browser on live video from the front-facing camera, displaying the labeled output in realtime.
Access to the front-facing camera is through the html5 WebRTC Media interface.
NOTE: The web app must be accessed over https in order for the browser to allow access to the camera.
Try it here https://www.isitredorgreen.com
If you want to run the python code, you'll need the following
- Python3 (https://www.python.org)
- OpenCV (https://www.opencv.org)
On MacOS:
% brew install python3
% pip3 install opencv-python