World-Model Image Scaling for Distance-Independent Object Recognition

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Image recognition systems struggle to accurately recognize objects at different scales in images captured from varying distances, requiring extensive training data and computational resources, which is particularly challenging in autonomous driving scenarios where objects like traffic lights need to be detected from close and far distances.

Innovation Solution

Utilizing ego vehicle localization and high-definition maps to rescale object views in images to a fixed scale, independent of capture distance, by determining the object's location in a world frame of reference and applying image scaling, enabling efficient image recognition with reduced training data and computational resources.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If image recognition is trained to recognize objects at different scales using multiple training images, then recognition accuracy across varying distances is improved, but training data requirements and computational resources increase significantly

Engineering Contradiction:
Improverecognition accuracyVSAvoidtraining data requirements
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

The patent changes the scale parameter of object views by rescaling images to a fixed scale based on world model information. This allows a single scale to represent multiple distance scenarios, eliminating the need for extensive multi-scale training data while maintaining recognition accuracy across varying distances

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent introduces a world model as an intermediary that provides prior knowledge about object locations and scales. This intermediary enables the system to infer appropriate rescaling parameters without requiring the image recognition component to learn all scale variations from training data

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If image recognition is trained to recognize objects at different scales using multiple training images, then recognition accuracy across varying distances is improved, but computational resources and model complexity increase

Engineering Contradiction:
Improverecognition accuracyVSAvoidmodel complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

By rescaling object views to a fixed scale using world model information, the patent reduces the parameter space that the image recognition model must handle. This allows for lighter-weight models with fewer parameters while maintaining accuracy, as the model only needs to recognize objects at one standardized scale rather than learning scale invariance

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent extracts the scale variation problem from the image recognition task by pre-rescaling images based on world model knowledge. This separates the scale normalization function from the recognition function, allowing the recognition model to focus solely on object identification without the added complexity of scale adaptation

Inventive Principle:
Principle #2Taking out (Extraction)

3Quantity of substance

If objects are captured at the same scale in images, then training data requirements are reduced, but the system cannot recognize objects from different distances

Engineering Contradiction:
Improvetraining data requirementsVSAvoiddistance range detection
Core Design Contradiction:
Quantity of substanceVSAdaptability or versatility

Solution Approach 1:

The patent performs preliminary rescaling of object views to a fixed scale using world model information before image recognition. This preliminary action normalizes all objects to a common scale, allowing the use of simplified training data while maintaining the ability to recognize objects from any distance through the pre-processing rescaling step

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent transitions from varying scale dimensions to a fixed scale dimension by rescaling based on world model parameters. This dimensional transformation allows the system to maintain adaptability across distance ranges while operating with a single scale dimension, effectively trading physical scale variation for computational scale normalization

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

Data Source

PatentEP4150516B1Processing images for extracting information about known objects for applying image scaling
Publication Date: 2025.10.29 FIVE AI LTD
  • EP4150516B1 patent drawingFigure 1
  • EP4150516B1 patent drawingFigure 2
  • EP4150516B1 patent drawingFigure 2

AI summary

A computer-implemented method of processing images for extracting information about known objects comprises the steps of receiving an image containing a view of a known object at a scale dependent on an object distance of the known object from an image capture location of the image; determining, from a world model representing one or more known objects in the vicinity of the image capture location, an object location of the known object, the object location and the image capture location defined in a world frame of reference; and based on the image capture location and the object location in the world frame of reference, applying image scaling to the image, to extract a rescaled image containing a rescaled view of the known object at a scale that is substantially independent of the object distance from the image capture location.