Image Registration Using Offset-Corrected Sensor Models

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image registration methods in remote sensing applications face inaccuracies due to errors in ground to image transformations and elevation data, leading to misalignment and distortion in sensed images, which complicates the comparison and fusion of images from different viewing geometries.

Innovation Solution

The method involves generating corrected ground to image transformations by applying offset parameters to the transformations, and using image comparison operations like phase correlation to determine the optimal corrections that maximize image similarity, thereby minimizing alignment and terrain perspective errors, and ensuring accurate registration of images in a common coordinate system.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If ground to image transformations are used to convert image coordinates to world coordinates and back, then image registration between different viewing geometries is achieved, but inaccuracies and errors in the transformation process lead to misalignment and distortion

Engineering Contradiction:
Improveimage registration accuracyVSAvoidtransformation error
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent applies feedback by using image comparison operations (such as phase correlation) to evaluate the alignment quality between registered images and iteratively adjust the ground to image transformations. The transformation parameters are refined based on the feedback from image similarity measurements, continuously improving registration accuracy while compensating for transformation errors.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent changes transformation parameters by applying offset corrections to the ground to image transformations. Multiple candidate transformations with different offset parameters are generated, and the optimal parameters are selected based on image comparison results, thereby adjusting the transformation to minimize misalignment and distortion.

Inventive Principle:
Principle #35Parameter changes

2Loss of information

If multiple sensed images from different viewing geometries are registered, then comprehensive spatial information is obtained, but the complexity of coordinate transformation and alignment increases

Engineering Contradiction:
Improvespatial information completenessVSAvoidregistration process complexity
Core Design Contradiction:
Loss of informationVSDevice complexity

Solution Approach 1:

The patent uses world coordinates as an intermediary to facilitate registration between images from different viewing geometries. By transforming all images to a common world coordinate system and then back to image coordinates, the intermediary enables comprehensive spatial information integration while managing the complexity of direct image-to-image alignment across different geometries.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent employs a universal ground to image transformation framework that can handle multiple images with different viewing geometries, sensors, and acquisition times. The same transformation process and image comparison operations are applied universally to all image pairs, simplifying the registration of multiple images while maintaining accuracy.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Ease of operation

If elevation information from Digital Elevation Models is used to constrain world coordinates, then image to ground coordinate conversion is enabled, but errors in elevation data propagate to transformation inaccuracies

Engineering Contradiction:
Improvecoordinate conversion feasibilityVSAvoidtransformation accuracy
Core Design Contradiction:
Ease of operationVSMeasurement precision

Solution Approach 1:

The patent performs preliminary actions by generating multiple candidate transformations with different offset parameters before final selection. This preliminary generation of transformation options allows the system to pre-compensate for potential elevation data errors by selecting the transformation that yields the best image alignment, rather than relying solely on the accuracy of the elevation model.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent uses feedback from image comparison operations to evaluate and select the optimal transformation among multiple candidates. The image similarity measurement provides feedback on which transformation best compensates for elevation data errors, allowing the system to overcome limitations in elevation model accuracy through empirical validation.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP3574472B1Apparatus and method for registering recorded images
Publication Date: 2021.09.29 THE SEC OF STATE FOR DEFENCE IN HER BRITANNIC MAJESTYS GOVERNMENT OF THE UK OF GREAT BRITAIN & NORTHERN IRELAND
  • EP3574472B1 patent drawingFigure 1~2
  • EP3574472B1 patent drawingFigure 3
  • EP3574472B1 patent drawingFigure 4

AI summary

A method and apparatus are provided for registering a base recorded image (9) of an object or terrain with a secondary recorded image (10) of the object or terrain, using a base image sensor model (12), secondary image sensor model (13) and elevation information (11). A plurality of biases (24) and respective matched biases (29) are applied to the secondary image sensor (model 13) and base image sensor model (12), to determine a plurality of corrected sensor models (26) and respective nominally corrected base image sensor models (31). Each corrected sensor model (26) and respective nominally corrected base image sensor model (31) is used to reproject (27) the secondary recorded image (10), and in each case the reprojected secondary recorded image is correlated (19) with the base recorded image (12) to ascertain a correlation score (20) and adjustments to the bias corrections (32). The correlation score being evaluated to determine optimally corrected sensor models and and/or an optimally transformed image.