Image Warping for Feature Identification in Dynamic Scenes

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image processing techniques struggle to accurately identify features within a scene due to difficulties in registering images captured at different points in time, especially when there is movement of the sensor or the underlying scene, leading to increased noise and decreased accuracy, particularly in real-time or near real-time applications.

Innovation Solution

The method involves obtaining a plurality of images captured at different points in time and space, defining a reference plane, applying image warping to align the images, and interpolating pixel coordinates to generate an interpolated representation of the image, which is then used to identify features within the scene, concurrently providing for rotation, translation, and warping in real-time or near real-time.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If images are registered to reduce noise and improve feature identification accuracy, then measurement precision is improved, but device complexity increases due to the need for registration algorithms and processing

Engineering Contradiction:
Improvefeature identification accuracyVSAvoidimage processing system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by pre-defining a reference plane and pre-computing transformation parameters before processing actual images. The reference plane is established in advance with known coordinate system, allowing rapid registration of incoming images without complex real-time calculations, thus reducing computational complexity while maintaining registration accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent introduces a reference plane as an intermediary element between the sensor plane and the feature identification process. This reference plane serves as a mediator that simplifies the registration process by providing a stable coordinate framework, reducing the complexity of direct image-to-image registration while improving measurement precision.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If higher order corrections such as rotation and nonlinear stretching are applied to register images, then measurement precision is improved, but productivity decreases due to computational intensity and inability to process in real-time

Engineering Contradiction:
Improveimage registration accuracyVSAvoidreal-time processing capability
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent applies parameter changes by transforming the registration problem from complex higher-order geometric transformations to simpler parameter adjustments. By changing the reference frame and using affine transformations with fewer parameters, the system achieves sufficient registration accuracy without the computational burden of full nonlinear warping, enabling real-time processing.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent applies partial action by implementing only the necessary registration corrections rather than complete higher-order transformations. The system applies sufficient registration to achieve the required measurement precision while omitting computationally intensive components, thereby maintaining real-time productivity.

Inventive Principle:
Principle #16Partial or excessive action

3Reliability

If images captured at different points in time are processed to identify moving objects, then reliability is improved through background elimination, but device complexity increases due to multiple image processing steps

Engineering Contradiction:
Improveobject detection reliabilityVSAvoidimage processing workflow complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent merges multiple image processing operations into a unified registration framework. By combining background elimination, object detection, and image registration into a single integrated process based on the reference plane, the system reduces workflow complexity while maintaining the reliability benefits of multi-step processing.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP4517680A1Image processing method and system for identifying a feature within a scene
Publication Date: 2025.03.05 THE BOEING CO
  • EP4517680A1 patent drawingFigure 1
  • EP4517680A1 patent drawingFigure 2~3
  • EP4517680A1 patent drawingFigure 4

AI summary

A computing device, method and computer program product are provided to identify a feature. In a method, images are captured at different points in time and space and a reference plane is defined in a reference frame. The method includes applying image warping to map a plurality of reference points of a respective image in a sensor plane to corresponding reference points of the reference plane and mapping pixel coordinates for a plurality of pixels of the reference plane to the respective image based upon a positional relationship between the reference plane and a representation of the respective image established by having applied image warping to the respective image. The method further includes interpolating from pixels of the respective image to pixels at the pixel coordinates of the reference plane to generate an interpolated representation of the respective image that is then used to identify the feature within the scene.