Automated Ground Control Point Selection for Satellite Image Correction

Solution Overview

Problem

Current methods for generating ground control points for high-definition digital maps are time and money intensive, as they require manual collection or human labeling, and there is a need for an automated approach to select and correct feature points for accurate camera model refinement in satellite imagery.

Innovation Solution

A system that automatically identifies and recommends ground control points by collecting images, detecting candidate feature points, performing feature correspondence, triangulating locations, and filtering points for optimal geographic distribution to correct satellite image errors, using machine learning and computer vision techniques.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual collection or human labeling methods are used to generate ground control points, then the accuracy of camera models can be improved, but the time and cost requirements increase significantly

Engineering Contradiction:

Improvecamera model accuracyVSAvoidtime for generating ground control points

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system automatically detects candidate feature points in satellite images and performs self-validation through geometric consistency checks across multiple images. The algorithm independently identifies, triangulates, and filters ground control points without requiring manual intervention, enabling the system to serve itself in the ground control point generation process.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent replaces the mechanical human labeling process with an automated computer vision system that uses image processing algorithms to detect feature points, perform feature correspondence, and triangulate locations. This substitution eliminates manual labor while maintaining the ability to generate accurate ground control points for camera model refinement.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If manual collection or human labeling methods are used to generate ground control points, then the accuracy of camera models can be improved, but the cost requirements increase significantly

Engineering Contradiction:

Improvecamera model accuracyVSAvoidcost for generating ground control points

Core Design Contradiction:

Measurement precisionVSLoss of energy

Solution Approach 1:

The system automatically detects candidate feature points in satellite images and performs self-validation through geometric consistency checks across multiple images. The algorithm independently identifies, triangulates, and filters ground control points without requiring manual intervention, enabling the system to serve itself in the ground control point generation process.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent replaces the mechanical human labeling process with an automated computer vision system that uses image processing algorithms to detect feature points, perform feature correspondence, and triangulate locations. This substitution eliminates manual labor while maintaining the ability to generate accurate ground control points for camera model refinement.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Productivity

If automated methods are used to select ground control points, then the time and cost efficiency can be improved, but the measurement precision may be compromised

Engineering Contradiction:

Improveefficiency of ground control point selectionVSAvoidaccuracy of selected feature points

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The system performs geometric consistency checks by triangulating the locations of candidate feature points across multiple satellite images and verifying their spatial coherence. This feedback mechanism validates the accuracy of automatically detected points, ensuring that only geometrically consistent points are selected as ground control points, thereby maintaining measurement precision while achieving automation.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent performs preliminary filtering of candidate feature points based on their detectability and geographic distribution before final selection. By pre-processing and validating candidate points through geometric consistency checks and triangulation, the system ensures high-quality ground control points are selected automatically without compromising accuracy.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS10984552B2Method, apparatus, and system for recommending ground control points for image correction

Publication Date: 2021.04.20 HERE GLOBAL BV

AI summary

An approach is provided for automatically recommending ground control points or any other feature points for image correction (e.g., satellite image correction). The approach, for example, involves collecting a plurality of images depicting a geographic area of interest. The approach also involves processing the plurality of images to detect one or more candidate feature points (e.g., ground control points or other features detectable in the images). The approach further involves performing a feature correspondence of the one or more candidate feature points across the plurality of images. The approach also involves triangulating respective locations of the one or more candidate feature points based on the feature correspondence. The approach further involves filtering the one or more candidate feature points based on the respective locations. The approach further involves providing the filtered one or more candidate feature points as an output comprising one or more recommended feature points.