PerceptIQ enables fully automated image acquisition through a synchronised multi-camera system designed for production-line integration. As each shoe passes through the inspection zone, a laser tripwire precisely detects its position and triggers simultaneous image capture across multiple viewpoints (top, left, and right). This ensures comprehensive visual coverage of complex geometries and curved surfaces in a single pass. By eliminating manual handling and standardising capture timing and positioning, the system produces consistent, high-quality input data while maintaining real-time throughput on conveyor systems.
Accurate colour evaluation requires strict control over environmental conditions. PerceptIQ addresses this through a custom-built lightbox that replicates industry-standard lighting environments using D65 (daylight) and TL84 (retail) fluorescent sources. These high-CRI light sources are configured to provide uniform, top-down illumination, minimising shadows, reflections, and colour distortion. The enclosed setup, enhanced with blackout curtains, isolates the system from ambient light interference, ensuring that captured images remain consistent across time and deployment locations. This controlled calibration is critical for producing reliable and repeatable colour measurements.
At the core of the pipeline is a robust, interpretable colour analysis framework grounded in colour science. The system first applies a trained segmentation model to isolate the shoe region at pixel-level precision, removing background noise and ensuring accurate colour extraction. It then computes the mean L*, a*, b* values of the segmented region and compares them against a reference standard using the ΔE metric, which quantifies perceptual colour difference. By directly measuring deviation rather than relying solely on learned predictions, this approach provides stable, explainable pass/fail decisions aligned with industrial tolerances, while maintaining near real-time performance.
