RGB vs HSV: making industrial color sorting more reliable

Color sorting sounds simple until changing illumination and reflective surfaces cause one product to receive different classifications.

Reproducible demo

OpenCV — Thresholding Operations using inRange provides Python/C++ examples with live HSV sliders. This is an algorithm demonstration, not a commissioned production-line video.

RGB versus HSV

RGB encodes red, green and blue channel intensity; all three respond to illumination.

HSV separates hue, saturation and value. With controlled light and sufficiently saturated colors, hue thresholds can be easier to tune.

For standard 8-bit OpenCV HSV data, Hue is normally encoded as 0–179. Since red wraps around the hue boundary, red detection often uses two threshold bands.

A practical pipeline

1. Stabilize lighting, product positioning and white balance.
2. Convert camera BGR to HSV with cv2.cvtColor().
3. Apply cv2.inRange() for target color ranges.
4. Clean the mask and inspect relevant contours or an ROI.
5. Reject ambiguous cases instead of forcing every image into a class.

import cv2

frame = cv2.imread("sample.jpg")
hsv = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV)
low_red = cv2.inRange(hsv, (0, 80, 60), (10, 255, 255))
high_red = cv2.inRange(hsv, (170, 80, 60), (179, 255, 255))
red_mask = cv2.bitwise_or(low_red, high_red)

These are illustrative, not validated factory thresholds.

Failure cases

Hue is unstable near zero saturation. Glare, shadows, mixed-spectrum lighting and automatic camera settings can break segmentation. For calibrated color judgement, consider controlled illumination, a reference chart and potentially Lab color-space measurements.

Source: OpenCV — inRange HSV tutorial.

Independent technical analysis; demos and media belong to their cited sources.

Cover image: OpenCV red-color segmentation result from Rajasekhar Guptha's tutorial; this is an algorithm demo, not a production line.