Multi-Camera Hyperspectral 3D Point Clouds for Camouflaged Object Detection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing three-dimensional point cloud generation methods using LiDAR and RADAR sensors are unreliable, especially in rainy conditions and fail to distinguish objects from backgrounds with similar colors, particularly in environments where objects are camouflaged.

Innovation Solution

Utilizing hyperspectral images from multiple cameras to determine three-dimensional positions of regions, applying geometric rules, neural networks, or unsupervised learning to match pixel signatures, and constructing a three-dimensional point cloud based on these positions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If LiDAR sensors are used to generate three-dimensional point clouds, then the generation speed is fast, but the reliability deteriorates in rainy conditions

Engineering Contradiction:
Improvepoint cloud generation speedVSAvoidsensor reliability in rain
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent changes the sensing parameter from optical (LiDAR) to electromagnetic spectrum-based (hyperspectral imaging) to maintain operation in adverse weather conditions. Hyperspectral cameras capture reflected light across multiple spectral bands, which can penetrate rain and fog better than traditional LiDAR, thus improving reliability while maintaining operational speed.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent uses composite sensing approaches by combining hyperspectral imaging data from multiple cameras with different spectral sensitivities. This composite approach creates a more robust point cloud generation system that maintains reliability in rainy conditions by leveraging the complementary strengths of different spectral bands.

Inventive Principle:
Principle #40Composite materials

2Ease of manufacture

If RGB images are used to create point clouds, then the processing is simple and fast, but the ability to distinguish objects from backgrounds deteriorates when colors are similar

Engineering Contradiction:
Improveprocessing simplicityVSAvoidobject-background distinction accuracy
Core Design Contradiction:
Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent transitions from three-color RGB imaging to hyperspectral imaging that captures data across dozens to hundreds of spectral bands. This dimensional expansion in the spectral domain provides additional information to distinguish objects from backgrounds with similar colors, as materials have unique spectral signatures that extend beyond visible colors.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

The patent changes the imaging parameter from three broad color channels (RGB) to numerous narrow spectral bands across the electromagnetic spectrum. This parameter change enables material identification based on spectral characteristics, dramatically improving object-background distinction while maintaining automated processing capabilities.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4645260A1Constructing a three-dimensional point cloud
Publication Date: 2025.11.05 BAE SYSTEMS PLC
  • EP4645260A1 patent drawingFigure 1
  • EP4645260A1 patent drawingFigure 2
  • EP4645260A1 patent drawingFigure 3~4

AI summary

The present disclosure relates to a computer-implemented method of constructing a three-dimensional point cloud of a real-world environment in which an entity operates. The computer-implemented method comprises: receiving a first hyperspectral image from a first hyperspectral camera, and a second hyperspectral image from a second hyperspectral camera, wherein the first and second hyperspectral cameras are mounted to the entity; determining positions of regions in the first hyperspectral image and positions of the regions in the second hyperspectral image; determining three-dimensional positions associated with each region based on the respective positions of the regions in each of the first and second hyperspectral images, a separation distance between the first and second hyperspectral cameras, and an orientation angle of each of the first and second hyperspectral cameras; and constructing the three-dimensional point cloud to include a point at each determined three-dimensional positions.