Multispectral imaging: see beyond human vision

AI-driven spectral imaging transforms data into actionable insights for your business.

Spectral imaging systems with active LED illumination
Image courtesy of MATIS

Multispectral imaging is a cutting-edge technology that captures data across multiple wavelengths of light, enabling detailed analysis beyond human vision. This advanced imaging technique is widely used in various industries, including manufacturing, art authentication, precision agriculture, food safety and processing, pharmaceuticals, and environmental monitoring. When combined with AI-powered data pipelines, multispectral imaging delivers rapid, non-destructive insights, making it an invaluable tool for industrial applications.

Understanding multispectral technology: From light acquisition to AI-powered analysis

AI-driven multispectral image processing

Multispectral imaging provides enriched information to be processed by a dedicated AI data pipeline.

The field of multispectral imaging combines the principles of spectroscopy and imaging techniques:

  • Spectroscopy: Analyzes the interaction of light with matter to provide detailed information about the chemical composition of materials.
  • Imaging techniques: Enable visualization and mapping of chemical properties across different areas of a sample.

The process involves:

  1. Acquisition with a specialized system (active illumination, imagers, filters)
  2. Data fed to a deep neural network (DNN).
  3. Segmentation of compounds based on spectral response and chemometric analysis.

At CSEM, multispectral imaging goes beyond conventional 2D images to a more advanced structure called a data cube. This combines spatial pixel distribution with spectral information for each pixel, creating a comprehensive three-dimensional dataset.

Various optical and hardware techniques can be used to capture spectral data, such as hyperspectral imagers, active illumination, and microlens-based vignetting. Each approach comes with its own benefits and limitations, and a high-level radar plot can be used to compare and evaluate different system types.

CSEM's advanced multispectral imaging uses

We offer tailored solutions to address any application, finding an optimum balance between spectral and spatial resolution, and cycle time. Our embedded processing through artificial intelligence at the edge to allow for fast decision-making in any environment.

Revolutionizing steel tracking: AI-based identification solution for steel mills

Multispectral camera acquisition setup
SMS Concast uses a multispectral vision system co-developed by CSEM to automatically read ID numbers of steel billets in challenging conditions.

Zurich-based SMS Concast, a leader in steel manufacturing, faced a major challenge in identifying steel billets stored outdoors. Over time, rust, weather conditions, and pollution obscured the embossed serial numbers, making traditional Optical Character Recognition (OCR) ineffective.

To address this, CSEM developed an AI-powered multispectral vision system, leveraging machine learning algorithms and multi-wavelength imaging to extract serial numbers from degraded billets with exceptional precision. Operating at 99.8% accuracy, even in harsh industrial conditions, the system enhances traceability and minimizes human intervention.

Key benefits:

  • Automated billet identification, ensuring reliable traceability
  • High accuracy (99.8%) in demanding environments
  • Seamless integration into existing industrial processes

Contactless real-time spectral analysis of wastewater

Sewage system
Contactless real-time spectral analysis of wastewater

Traditional wastewater monitoring requires manual sampling and laboratory testing, leading to delays in detecting contamination and inefficiencies in water treatment.

CSEM’s real-time, contactless spectral imaging solution, powered by machine learning algorithms and spectral signature analysis, instantly extracts chemical composition data from water. By analyzing the diffuse reflectivity of water, the system provides continuous, maintenance-free monitoring.

Key indicators measured:

  • Turbidity (TUR): Predicts water transparency with 99% accuracy
  • Chemical Oxygen Demand (COD): Measures organic compound concentration with 93% accuracy

Impact:

  • Real-time detection of contamination
  • Autonomous inline monitoring with zero maintenance
  • Faster intervention, reducing environmental impact

New standards in art identification with multispectral imaging

Multispectral art analysis on a painting
Multispectral art analysis provides insights on the authenticity of paintings

Art authentication and restoration rely on visual inspection and conventional imaging techniques, which often fail to detect hidden details, pigment composition, and underlying sketches.

MATIS, a Swiss startup specializing in art preservation, partnered with CSEM to develop an AI-enhanced multispectral imaging system. This solution reveals hidden artwork layers, detects forgeries, and analyzes material composition using spectral analysis and AI-driven image processing.

Key benefits:

  • Non-invasive authentication of historical artwork
  • Enhanced pigment and material identification
  • Scientific validation for restoration and conservation

Multispectral sensors for skin and tissue segmentation for cancer diagnosis

CSEM’s multispectral system provides morphological and depth information in skin cancer analysis
High spatial resolution from CSEM’s multispectral system provides morphological and depth analysis in skin cancer analysis when combined with a convolutional neural network (CNN).

Detecting and analyzing skin and tissue abnormalities remains a major challenge in early cancer detection. Conventional imaging methods may lack the spatial and spectral resolution needed to differentiate between healthy and malignant tissues effectively.

Our handheld spectral imagers, SpectroX and SpectrHal, offer up to 25 bands based on active illumination, using low-power LEDs for targeted spectral selection. By integrating AI-driven algorithms, this system provides precise morphological and depth information, enhancing early diagnosis capabilities.

Key benefits:

  • Non-invasive and real-time detection of suspicious tissue structures
  • Cost-effective imaging solution with high spatial resolution
  • Enhanced cancer segmentation when combined with convolutional neural networks (CNNs)

Application of AI-based multispectral technique in food inspection

Multispectral camera system to suppress the ambient parasitic light effects.
Snapshot multispectral camera is combined with custom high-power illumination to suppress ambient parasitic light effects

Traditional food inspection methods rely on surface-level visual checks, missing contaminants, ripeness indicators, and adulteration.

CSEM’s high-speed multispectral imaging system, combined with AI-driven analysis, enables real-time assessment of food safety, freshness, and composition.

Key benefits:

  • Detects contamination, ripeness, and moisture content
  • Non-destructive, real-time assessment
  • Optimized for high-speed food processing lines
Logo Buhler Groupe

Mario Urbisaglia

Team Manager R&D

CSEM was the ideal partner for this project: expert in intelligent systems and algorithms, pragmatic and flexible when we had to change the data collection plan, and fast. This was our first project with CSEM—and it won’t be the last.

Partner with CSEM to unlock multispectral imaging’s full potential

CSEM’s AI-powered multispectral imaging solutions deliver unmatched precision, automation, and efficiency across manufacturing, agriculture, healthcare, and conservation. Our dedicated AI algorithms minimize hardware requirements, enabling lightweight, faster, cost-effective, on-site detection in diverse applications, including drones, high-speed production lines, and handheld battery-operated multispectral systems.

Discover the power of AI-driven multispectral imaging today!

Exploring the science behind spectral information

Interested in the latest research on multispectral imaging technology? Explore our publications below. Some articles may require a subscription or institutional access.