Infrared Pedestrian Detection Using Temperature-Selected AI Models

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Infrared imaging systems face challenges in pedestrian detection due to contrast inversion between pedestrians and surroundings based on weather conditions, leading to reduced accuracy in identifying targets with small temperature differences.

Innovation Solution

A method involving a set of machine learning models defined by temperature ranges, where each model is trained with classified infrared images, allowing for accurate detection by selecting the appropriate model based on the temperature characteristics of the target and comparison objects.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If a single machine learning model is used for pedestrian detection in infrared images, then the system complexity is low, but the detection accuracy deteriorates when temperature difference between target and background is small

Engineering Contradiction:

Improvesystem complexityVSAvoiddetection accuracy

Core Design Contradiction:

Device complexityVSMeasurement precision

Solution Approach 1:

The patent divides the detection system into multiple specialized machine learning models, each trained for specific temperature range conditions. This segmentation allows each model to specialize in detecting pedestrians under particular thermal contrast scenarios, thereby improving overall detection accuracy without requiring a single overly complex universal model.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system dynamically selects which machine learning model to use based on the current temperature characteristics of the infrared image. By adapting the model selection to the actual thermal conditions, the system maintains high detection accuracy across varying environmental conditions while avoiding the need for a static, overly complex model that must handle all scenarios.

Inventive Principle:

Principle #15Dynamics

2Adaptability or versatility

If machine learning models are trained with diverse temperature conditions, then the model can handle various weather scenarios, but the detection accuracy deteriorates for specific temperature ranges due to diluted specialization

Engineering Contradiction:

Improveweather condition adaptabilityVSAvoiddetection accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

Instead of training a single model on all temperature conditions, the patent segments the training data by temperature ranges and creates separate specialized models for each segment. This allows each model to achieve high accuracy for its specific temperature range while the collection of models provides broad weather adaptability through selective deployment.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Each machine learning model is optimized for specific local conditions (temperature ranges) rather than attempting to be universally optimal. This local specialization ensures that each model has superior detection accuracy for its designated temperature scenario, while the system as a whole maintains versatility through multiple specialized models.

Inventive Principle:

Principle #3Local quality

3Adaptability or versatility

If infrared images with inverted contrast are included in the training data set, then the training data diversity is high, but the model's ability to detect targets with small temperature differences deteriorates

Engineering Contradiction:

Improvetraining data diversityVSAvoidtarget detection accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent segments the training data based on temperature contrast characteristics and trains separate models for different contrast scenarios. This segmentation allows the system to maintain high detection accuracy for small temperature differences by using models specifically trained on such conditions, while still achieving broad adaptability through the collection of specialized models.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system changes the parameter of model selection based on the temperature characteristics of the input image. By adjusting which model is deployed according to the thermal contrast parameters, the system maintains high detection accuracy across diverse conditions without mixing contradictory training scenarios in a single model.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20250378685A1Method of training machine learning model for detecting target object in infrared image, method executed by processor therefor, onboard vehicle computer unit therefor, non-transitory computer readable storage medium comprising program codes therefor, vehicle

Publication Date: 2025.12.11 SUBARU CORP

AI summary

A method classifies an IR image to be used for training a machine learning model. The method comprises: (i) providing a set of machine learning models in function of temperature; (ii) acquiring an infrared image; (iii) identifying a target object to be detected and a comparison object in the infrared image; (iv) calculating a first characteristic of the target object in the infrared image and a second characteristic of the comparison object in the infrared image; and (v) based on the first characteristic of the target object and the second characteristic of the comparison object, selecting a machine learning model among the set of machine learning models in function of temperature for the IR image.