Machine Learning Device for Far-Infrared to Visible Light Image Conversion

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing techniques face challenges in accurately converting far-infrared images into visible light images, especially in transitioning from nighttime infrared images to daytime color images, due to limitations in specifying accurate color values for pixel values in infrared images.

Innovation Solution

A machine learning device is employed to generate a trained visible light image generation model by acquiring and processing far-infrared images from different time zones and corresponding visible light images. This model includes two components: a first generation model that converts nighttime far-infrared images into daytime far-infrared images, and a second generation model that converts daytime far-infrared images into daytime visible light images.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a far-infrared image is converted into a visible light image using existing techniques, then the image can be recognized through human eyes, but the accuracy of color value specification is limited and cannot accurately represent the original scene

Engineering Contradiction:

Improvecolor value accuracyVSAvoidcolor information loss

Core Design Contradiction:

Measurement precisionVSLoss of information

Solution Approach 1:

The patent introduces a far-infrared image captured at a predetermined time zone as an intermediary to bridge the gap between nighttime far-infrared images and daytime visible light images. This intermediary contains temporal and environmental information that helps the neural network accurately infer color values, thereby resolving the contradiction between improving color accuracy and preventing color information loss.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent performs preliminary capture of a far-infrared image at a predetermined time zone before the actual conversion process. This pre-captured image serves as training data and reference information for the neural network, enabling it to predict accurate color values during the conversion from nighttime far-infrared to daytime visible light images.

Inventive Principle:

Principle #10Preliminary action

2Difficulty of detecting and measuring

If a nighttime infrared image is converted into a daytime color image, then object detection becomes easier, but it is difficult to accurately represent daytime color information from nighttime infrared data

Engineering Contradiction:

Improveobject detection easeVSAvoidcolor information accuracy

Core Design Contradiction:

Difficulty of detecting and measuringVSManufacturing precision

Solution Approach 1:

The far-infrared image captured at a predetermined time zone acts as a mediator that contains environmental and temporal characteristics. This intermediary enables the neural network to learn the relationship between nighttime infrared scenes and daytime color appearances, making object detection easier while maintaining color information accuracy.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces traditional mechanical or algorithmic image conversion methods with a neural network-based machine learning system. This substitution enables the system to learn complex mappings from nighttime infrared images to daytime color images, overcoming the limitations of conventional techniques in accurately representing color information.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Device complexity

If a direct conversion method is used from nighttime far-infrared image to daytime visible light image, then the processing is simple, but the conversion accuracy is limited

Engineering Contradiction:

Improveconversion process simplicityVSAvoidconversion accuracy

Core Design Contradiction:

Device complexityVSMeasurement precision

Solution Approach 1:

The patent segments the conversion process into multiple components: acquiring nighttime far-infrared image, acquiring far-infrared image at predetermined time zone, and acquiring daytime visible light image. These segmented data sets are used to train the neural network, which then performs the actual conversion. This segmentation improves conversion accuracy while keeping the operational process relatively simple.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary data collection and neural network training using multiple image sets before the actual conversion operation. This preliminary action prepares the system with learned knowledge, enabling accurate conversion during actual use without requiring complex real-time processing.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS12267570B2Machine learning device, image processing device, machine learning method, and machine learning program

Publication Date: 2025.04.01 JVC KENWOOD CORP

AI summary

A visible light image generation model learning unit generates a trained visible light image generation model that generates a visible light image in a second time zone from a far-infrared image in a first time zone. The visible light image generation model learning unit includes a first learning unit that machine-learns the far-infrared image in the first time zone and a far-infrared image in the second time zone as teacher data and generates a trained first generation model that generates the far-infrared image in the second time zone from the far-infrared image in the first time zone, and a second learning unit that machine-learns the far-infrared image in the second time zone and the visible light image in the second time zone as teacher data and generates a trained second generation model that generates the visible light image in the second time zone from the far-infrared image in the second time zone.