Research

Spectrophotometers measure colour accurately, but they are expensive, read one small spot at a time and are used off-line. Manual inspection is fast but subjective. This system sits between the two: a webcam in a controlled lighting booth watches a large area of fabric and reports colour differences in real time.

Each frame is converted to CIELAB and the median colour of the inspection region is compared with a reference using CIEDE2000, the current CIE recommendation. Before any test, five checks must pass — camera ready, not too dark, not over-exposed, stable over time, and evenly lit across a 3 × 3 grid. A calibration step measures the system's own noise on a uniform fabric and sets the pass/fail limit above it, so the system never reports noise as a defect.

The software separates the measurement camera from any other camera: a laptop webcam still works for demonstrations, but its results are clearly marked as demo readings.

Bangladesh's textile sector still relies largely on manual inspection, which is estimated at only 60–75% accuracy. This quantitative cross-sectional study compared eight factories in Bangladesh — five IoT-enabled and three without IoT — through a survey of 80 quality management professionals, using the Technology Acceptance Model (TAM) and the Technology–Organisation–Environment (TOE) framework.

All six hypotheses were supported at p < 0.01. IoT-enabled factories reported markedly lower shade mismatch (Cohen's d = 4.11) and were 5.32 times more likely to report lower defect rates. Non-IoT firms were 12.86 times more likely to see implementation cost as a critical barrier, and digital training correlated strongly with user satisfaction (r = 0.826). These findings motivated the low-cost machine vision system above.