Camouflaged Object Classification With Hyperspectral Material Verification

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image classification methods, particularly using RGB images, struggle with false classifications due to camouflaged objects in real-world environments, making it difficult to accurately detect and verify objects.

Innovation Solution

Utilizing a hyperspectral imaging system to construct a semantic materials map by labeling each pixel with its hyperspectral signature, followed by clustering and verifying the classified object based on expected materials, enhancing classification accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If RGB image classification is used, then processing speed is maintained, but classification accuracy deteriorates due to camouflaged objects

Engineering Contradiction:
Improveclassification accuracyVSAvoidimaging system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent combines RGB imaging and hyperspectral imaging into a dual-camera system. The RGB camera provides standard visual information while the hyperspectral camera captures material-specific spectral signatures. By merging these two imaging modalities, the system achieves both high classification accuracy through material verification and maintained processing speed through efficient data fusion techniques.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent introduces a semantic materials map as an intermediary structure that bridges RGB image classification and hyperspectral material verification. This map stores verified material information from hyperspectral data and uses it to correct or validate RGB-based classifications, effectively mediating between the two imaging systems to improve accuracy without requiring full hyperspectral processing for all objects.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If hyperspectral imaging is used for material identification, then detection capability improves for camouflaged objects, but processing time increases

Engineering Contradiction:
Improveobject detection reliabilityVSAvoidprocessing time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent performs preliminary classification using the RGB image and trained classifier before conducting detailed material verification with hyperspectral data. Objects that can be confidently identified through RGB imaging alone are processed quickly without hyperspectral analysis. Only objects requiring material verification (such as camouflaged objects or those with ambiguous RGB classifications) undergo the time-consuming hyperspectral analysis, thus reducing overall processing time while maintaining high detection reliability.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent applies partial hyperspectral analysis by using the semantic materials map to verify only the material composition of suspected objects rather than analyzing all pixels in the entire image. This selective approach allows the system to maintain high reliability for critical detections while minimizing processing time by avoiding unnecessary hyperspectral processing for clearly identifiable objects.

Inventive Principle:
Principle #16Partial or excessive action

3Difficulty of detecting and measuring

If edge detection is performed on camouflaged objects, then object localization is attempted, but detection capability deteriorates due to background blending

Engineering Contradiction:
Improveobject detection difficultyVSAvoidobject detection accuracy
Core Design Contradiction:
Difficulty of detecting and measuringVSMeasurement precision

Solution Approach 1:

The patent exploits color and spectral changes at the material level rather than relying solely on visual color and edge detection. By capturing hyperspectral signatures across multiple frequency bands (visible, infrared, ultraviolet), the system can detect material composition changes that are not visible to the human eye or standard RGB cameras. This allows accurate detection of camouflaged objects through their unique spectral fingerprints even when their visual appearance blends with the background.

Inventive Principle:
Principle #32Color changes

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

Improves object classification accuracy by distinguishing materials through hyperspectral signatures, effectively identifying and validating objects, even when camouflaged, thereby improving autonomous system operations.

Implementation Method 1

a hyperspectral image of the real-world environment from a hyperspectral camera mounted to the entity

Methodology Applied
Scientific EffectHyperspectral imaging: Absorption Spectroscopy

Data Source

PatentEP4645257A1Classifying an object
Publication Date: 2025.11.05 BAE SYSTEMS PLC
  • EP4645257A1 patent drawingFigure 1
  • EP4645257A1 patent drawingFigure 2~3
  • EP4645257A1 patent drawingFigure 4

AI summary

The present disclosure relates to a computer-implemented method of classifying an object from a real-world environment in which an entity operates. The computer-implemented method comprises: receiving an image of the real-world environment from a camera mounted to the entity, and a hyperspectral image of the real-world environment from a hyperspectral camera mounted to the entity; inputting the image to a trained image classifier to classify an object from the image; constructing a semantic materials map of the real-world environment by labelling each pixel of the hyperspectral image with a material based on its hyperspectral signature; identifying the classified object in the semantic materials map; and verifying the classified object as one of valid and invalid based on whether the material of the object in the semantic materials map matches an expected material for the classified object.