Camera Color Calibration for Distinguishing Red and Yellow Traffic Lights

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing camera calibration methods, particularly for autonomous vehicles, struggle to accurately distinguish between similar color spectrums like red and yellow, and traditional color correction matrices (CCMs) fail to account for actual spectral colors of traffic lights, leading to inaccurate RGB value outputs.

Innovation Solution

A supplemented Macbeth chart with additional color patches, determined using a spectrometer, is used to calibrate cameras, and real-time adjustments are made using HSV representation to optimize color correction matrices (CCMs).

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a traditional Macbeth chart with 24 color patches is used for camera calibration, then the calibration process is simple and quick, but the breadth of colors that can be calibrated is limited and similar color spectrums (like red and yellow) cannot be distinguished accurately

Engineering Contradiction:
Improvecolor detection accuracyVSAvoidcalibration target complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The calibration chart is segmented into multiple patches, each representing a specific traffic light color (red, yellow, green) with unique spectral characteristics. This segmentation allows the camera to learn distinct spectral signatures for each color, improving differentiation between similar colors while maintaining a manageable chart structure.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The calibration process transitions from traditional 2D RGB color space to 3D spectral space by incorporating spectral reflectance measurements across multiple wavelengths (400-1000nm). This dimensional expansion enables the system to distinguish between colors that appear similar in RGB but have different spectral compositions, such as red and yellow traffic lights.

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

2Measurement precision

If traditional Color Correction Matrices (CCMs) are used, then the processing is computationally efficient, but they fail to account for actual spectral colors of traffic lights leading to inaccurate RGB outputs

Engineering Contradiction:
ImproveRGB value accuracyVSAvoidsignal processing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system changes the parameters used for color correction from traditional 3x3 CCM operating in RGB space to spectral-based correction matrices that operate in multi-dimensional spectral space. This allows the system to account for the actual spectral characteristics of traffic light colors, improving accuracy while the dimensionality reduction technique keeps computational complexity manageable.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

A spectral library acts as an intermediary between the raw spectral measurements and the final color correction. This library contains pre-characterized spectral signatures of traffic light colors, serving as a reference that mediates the conversion from spectral data to accurate RGB outputs, improving both accuracy and computational efficiency.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Measurement precision

If spectral measurements are incorporated into calibration, then color spectrum distinction improves, but the calibration process becomes more complex and time-consuming

Engineering Contradiction:
Improvespectral color distinctionVSAvoidcalibration time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The spectral characteristics of traffic light colors are pre-measured and stored in a spectral library before actual camera calibration. This preliminary action allows the calibration process to reference pre-characterized data rather than performing full spectral analysis during calibration, maintaining high spectral distinction accuracy while significantly reducing calibration time.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

Instead of performing complex spectral measurements during actual camera calibration, the system creates a spectral library copy of traffic light color characteristics that can be referenced during calibration. This copying approach preserves the detailed spectral information needed for accurate color distinction while avoiding the time-consuming nature of real-time spectral measurement during calibration.

Inventive Principle:
Principle #26Copying

Data Source

PatentEP3985963B1Systems and methods for improved camera color calibration
Publication Date: 2025.10.29 ARGO AI LLC
  • EP3985963B1 patent drawingFigure 1
  • EP3985963B1 patent drawingFigure 2
  • EP3985963B1 patent drawingFigure 3

AI summary

Systems, methods, and computer-readable media are disclosed for improved camera color calibration. An example method may involve capturing a first wavelength emitted by a first type of traffic light. The example method may also involve determining, based on the first wavelength, a first color value associated with the wavelength emitted by the first type of traffic light. The example method may also involve capturing, by a first camera, a first image, video, or real-time feed of a first portion of a test target, the first portion of the test target including a first light color that is based on the first color value. The example method may also involve determining, based on the first image, video, or real-time feed of the first portion of a test target, a second color value output by the camera. The example method may also involve determining, based on a comparison between the first color value and the second color value, that a difference exists between the first color value and the second color value. The example method may also involve adjusting, based on the detection that a difference exists between the first color value and the second color value, at least one of: a property of the camera or signal processing associated with the camera.