Real-Time Hyperspectral Classification of Visually Similar Materials: A Pilot Study

10 pages•Published: June 2, 2026

Rana Muhammad Irfan Anwar and Eric M. Wetzel

Abstract

Visual inspection and Red, Green, Blue (RGB) cameras remain the dominant method for assessing surface finish and quality in construction materials. However, this approach is inherently subjective and constrained by human color perception, particularly with visually similar coatings. Similarly, conventional RGB imaging also fails to capture the subtle spectral cues that distinguish surface finishes sharing similar color and texture, leading to potential inaccuracies when verifying completed work. Reliable differentiation of surface coatings is essential for automated progress tracking applications. To address these limitations, this pilot study investigates the use of Hyperspectral Imaging (HSI) for the non-destructive differentiation of coated and uncoated gypsum surfaces under controlled lighting. Experiments were conducted in a laboratory under controlled illumination as a pilot study, using a Kelvin Play LED source. The lighting temperature varied from 6000 K to 8000 K to improve spectral separability. Data was collected as hyperspectral files in JSON format and applied Min-max normalization. Two classifiers, Support Vector Machine (SVM) and Random Forest Classifier (RFC), were trained and evaluated using spectra. The RFC achieved over 95% accuracy in real-time classification under 8000 K illumination. The live system, implemented via the Living Optics SDK (v1.9.0), predicted surface types directly on a grayscale camera feed with OpenCV overlays. The results confirm that hyperspectral sensing, coupled with optimized lighting and machine learning, can enable reliable, real-time differentiation of construction surface materials. The findings establish a strong foundation for extending hyperspectral inspection to mobile robotic platforms for autonomous, on-site progress tracking.

Keyphrases: hyperspectral imaging, machine learning, progress tracking

In: Wesley Collins, Anthony Perrenoud and John Posillico (editors). Proceedings of Associated Schools of Construction 62nd Annual International Conference, vol 7, pages 595-604.

Links:	https://easychair.org/publications/paper/nzqV
	https://doi.org/10.29007/gcqz

BibTeX entry

@inproceedings{ASC2026:Real_Time_Hyperspectral_Classification,
  author    = {Rana Muhammad Irfan Anwar and Eric M. Wetzel},
  title     = {Real-Time Hyperspectral Classification of Visually Similar Materials: A Pilot Study},
  booktitle = {Proceedings of Associated Schools of Construction 62nd Annual International Conference},
  editor    = {Wesley Collins and Anthony Perrenoud and John Posillico},
  series    = {EPiC Series in Built Environment},
  volume    = {7},
  publisher = {EasyChair},
  bibsource = {EasyChair, https://easychair.org},
  issn      = {2632-881X},
  url       = {/publications/paper/nzqV},
  doi       = {10.29007/gcqz},
  pages     = {595-604},
  year      = {2026}}

Download PDF Open PDF in browser