Computer vision basics

AdvancedAdvanced45 minBefore this: Cameras as sensors, MicroPython basics

What you’ll learn

How a computer stores an image
Finding a coloured object with thresholding in HSV
Contours, and the step up to AI object detection

Images are grids of numbers

A colour image is a grid of pixels, and each pixel has three numbers (blue, green and red in OpenCV’s order), each 0–255. A 640 × 480 image is 307,200 pixels.

HSV is better for colour

Lighting changes RGB values a lot. HSV separates Hue (the colour itself), Saturation (how strong) and Value (how bright), so “red-ish” stays in a narrow hue range even as light changes.

Finding a coloured object

Convert the image to HSV.
Threshold: keep only pixels within a hue/saturation/value range → a black-and-white mask.
Clean the mask (erode/dilate to remove speckles).
Find contours (outlines of white areas) and take the biggest.
Its centre gives the object’s position; its area hints at distance.

Beyond colour

To recognise things (people, cups, signs) robots use trained neural networks (for example YOLO-family detectors) that output boxes and labels. These need more computing power: a Raspberry Pi 5 can run small models, and AI accelerators make it faster (see AI on small boards).

Find a red ball (Python + OpenCV)

import cv2
import numpy as np

cam = cv2.VideoCapture(0)
while True:
    ok, frame = cam.read()
    if not ok:
        break
    hsv = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV)
    # red wraps around hue 0, so use two ranges
    mask = cv2.inRange(hsv, (0, 120, 70), (10, 255, 255)) | cv2.inRange(hsv, (170, 120, 70), (180, 255, 255))
    mask = cv2.erode(mask, None, iterations=2)
    mask = cv2.dilate(mask, None, iterations=2)
    contours, _ = cv2.findContours(mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
    if contours:
        c = max(contours, key=cv2.contourArea)
        (x, y), r = cv2.minEnclosingCircle(c)
        if r > 10:
            cv2.circle(frame, (int(x), int(y)), int(r), (0, 255, 0), 2)
            print("ball at x =", int(x))     # steer to keep x near the centre
    cv2.imshow("view", frame)
    if cv2.waitKey(1) == 27:   # Esc quits
        break

Steering from the result

error = x - frame_width / 2, then turn proportionally, as in Feedback and PID.

Finished this lesson?