Heterogeneous Image Change Detection for Disaster Damage Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting disaster damage using heterogeneous image data, such as those from satellites and unmanned aerial devices, face challenges due to differences in sensor types and lack of reference images, leading to inaccurate change detection in damaged areas.
Innovation Solution
A method and apparatus that process observation and reference images based on event and device types, preprocess the images to exclude exceptional objects, and apply change vector analysis (CVA) and principal component analysis (PCA K-means) to accurately detect changes in disaster-affected regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If satellite images are used for change detection based on predetermined tracks and periods, then wide area damage information can be efficiently collected, but damaged areas may not be captured at necessary times or may not belong to satellite regions
Solution Approach 1:
The system segments the monitoring approach by dividing it into satellite-based wide-area periodic monitoring and drone-based targeted on-demand monitoring. Satellite images provide broad coverage while drone images provide detailed, timely observations of specific damaged areas, resolving the contradiction between wide coverage and precise timely detection.
2Measurement precision
If unmanned aerial devices are used to approach disaster areas, then disaster-related data can be obtained at necessary times, but there is no reference image for detecting damaged areas
Solution Approach 1:
The system performs preliminary action by collecting and storing satellite images before disasters occur as reference images. When disasters happen, these pre-collected reference images are available for comparison with post-disaster drone images, enabling accurate change detection without needing to capture pre-disaster drone footage.
3Adaptability or versatility
If images from unmanned aerial devices and satellite images are combined for change detection, then both wide coverage and detailed observation are achieved, but errors occur due to different sensor attributes
Solution Approach 1:
The system introduces an intermediary normalization process that standardizes images from different sensors (satellite and drone) into a common reference framework. This intermediary step converts heterogeneous image data with different attributes into comparable formats, enabling accurate change detection while maintaining flexibility in data sources.
4Device complexity
If change detection is performed without considering disaster type, then processing is simplified, but accuracy of change detection is significantly degraded
Solution Approach 1:
The system implements dynamic processing complexity by automatically selecting disaster types and adjusting processing parameters based on the specific situation. The complexity adapts to each case - simple cases use basic processing while complex cases use advanced disaster-type-specific algorithms, optimizing both efficiency and accuracy.
Data Source
AI summary
Disclosed herein method and apparatus for detecting changes between heterogeneous image data for identifying disaster damage.The method includes obtaining observation image data of an event occurrence region and reference image data that occurs before an event and has a heterogeneous type of an image from the observation image data; processing the observation image data and the reference image data according to an attribute of the observation image data and preprocessing the observation image data and the reference image data to determine whether or not there is an exceptional object to be excluded from a change analysis in the processed observation image data and the processed reference image data based on an event type and to process the exceptional object; and detecting a change area of the event occurrence region based on the preprocessed observation image data and the preprocessed reference image data.