Computer-Aided Remote Sensing Image Interpretation with Likelihood-Based Modification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for remote sensing image interpretation face challenges such as high ambiguity in human visual interpretation, increased workload due to large image areas, and inconsistency in automatic classification results, especially with varying sensor characteristics and weather conditions.
Innovation Solution
A computer-aided image interpretation method that correlates feature positions and types with their likelihoods, allowing for the identification of features with high variability or incorrect interpretation results, facilitating their modification and improving accuracy through automatic classification and data storage in a device comprising a memory unit, processing unit, and display unit.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automatic classification is used to replace manual interpretation, then productivity increases, but measurement precision deteriorates due to ambiguity and inconsistency in classification results
Solution Approach 1:
The system calculates similarity between automatic classification results and manual interpretation results, then uses this feedback to identify and prioritize features requiring re-interpretation. This feedback mechanism allows the system to learn from discrepancies and improve accuracy while maintaining high productivity through selective human review.
Solution Approach 2:
The interpretation process is segmented into automatic classification and selective manual verification. Instead of requiring complete manual interpretation of all features, the system segments the workload to only those features with low similarity scores, thereby maintaining productivity while improving precision where needed.
2Measurement precision
If manual interpretation is performed to ensure accuracy, then measurement precision improves, but productivity deteriorates due to the large area of remote sensing images
Solution Approach 1:
Instead of performing complete manual interpretation of all features in large-area images, the system applies partial manual action only to features with low similarity between automatic and manual results. This selective approach maintains high accuracy for critical features while preserving overall productivity.
3Measurement precision
If multiple remote sensing images from different satellites are used to improve accuracy, then measurement precision improves, but device complexity increases due to data management requirements
Solution Approach 1:
The similarity calculation mechanism serves multiple functions: it compares automatic and manual results, identifies features requiring re-interpretation, and can be applied across multiple satellites and image types. This universal approach manages complexity by using a single consistent method across diverse data sources.
4Measurement precision
If spatial resolution is improved to enhance feature interpretation accuracy, then measurement precision improves, but loss of time increases due to larger image display areas
Solution Approach 1:
The system extracts only the essential comparison metric (similarity score) from the complex image interpretation process. By taking out this key indicator, the system avoids the time-consuming process of manually reviewing entire high-resolution images while still achieving accurate feature identification through targeted similarity-based selection.
Data Source
AI summary
A computer-aided image interpretation method and a device thereof to easily obtain an accurate image interpretation result are provided. An automatic classification means of the image interpretation device performs automatic classification by one of spectral characteristics, radiometric characteristics, diffuse characteristics, textures and shapes, or combinations thereof and accumulates data to an interpretation result database, for plural features of the same kind obtained by interpreting a remote sensing image obtained with an observation sensor. A means for extracting candidate of modification of interpretation result extracts the candidate of modification of interpretation result by comparing likelihoods that are the automatic classification results. A reinterpretation is performed for the candidate of modification of interpretation, and an interpretation result database is updated by an interpretation result update means. As a result, modification of the interpretation work can be efficiently performed.