Model-Based Image Change Quantification for Geospatial Imagery
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in accurately detecting and quantifying changes in geospatial imagery, particularly due to factors like shadows and obstructions, which can lead to uncertainty in feature extraction and change detection.
Innovation Solution
The Model-Based Change Quantification (MCQ) system addresses this by segmenting images using a scene model, extracting features while accounting for shadows and obstructions, and quantifying changes based on differences between feature vectors factoring in uncertainties.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional change detection methods are used without accounting for shadows and obstructions, then the processing speed is faster, but the measurement precision of feature extraction deteriorates due to uncertainty
Solution Approach 1:
The patent segments the image into multiple regions based on a scene model, separating areas affected by shadows and obstructions from those that are not. This allows the system to process different regions with different levels of detail and uncertainty, improving feature extraction precision without requiring complete reprocessing of the entire image.
Solution Approach 2:
The patent introduces a scene model as an intermediary that represents the expected appearance of objects and their relationships. This scene model serves as a mediator between the raw image data and the change detection process, allowing the system to account for shadows and obstructions by comparing actual image features against the modeled expectations.
2Reliability
If features are extracted without accounting for shadows and obstructions, then the extraction process is simpler, but the reliability of change detection deteriorates
Solution Approach 1:
The patent applies local quality by treating different regions of the image differently based on their characteristics. Areas with shadows or obstructions are marked with higher uncertainty weights, while clear areas use standard weighting. This allows the extraction process to maintain high reliability by adapting to local conditions without requiring complete redesign of the entire extraction process.
Solution Approach 2:
The patent incorporates feedback by using the scene model to continuously refine feature extraction and change detection. The modeled scene provides expected feature values that feedback into the extraction process, allowing the system to adjust and improve reliability by comparing actual extracted features against the modeled expectations and correcting for shadow and obstruction effects.
3Measurement precision
If a scene model is used to segment images and extract features, then the accuracy of change quantification improves, but the computational time increases
Solution Approach 1:
The patent applies partial action by extracting and processing only the features and regions that are most relevant for change detection, rather than processing the entire image uniformly. By using the scene model to identify and prioritize critical features, the system achieves high accuracy change quantification while reducing computational time by focusing resources on the most informative areas.
Data Source
AI summary
A system is provided for identifying and quantizing a change in a scene having objects based on comparison of features representing a target object derived from a first image and a second image that includes the target object. The system segments, based on a scene model of the scene, a second image that includes the target object to identify a second area representing the target object within the second image. The system extracts, from the second image, second features for the target object based on the second area. The system determines a second effectiveness of the second features indicating effectiveness of the second features in representing the target object. The system detects a change based on a difference between first features and the second features factoring in a first effectiveness and the second effectiveness.