Change Detection for Image Data Freshness
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Maintaining maximum coverage of high-resolution aerial and street-level images is expensive, leading to outdated online image data, as frequent acquisition of high-quality images is costly and storage limitations in vehicles restrict data upload.
Innovation Solution
Acquire low-resolution aerial images, compare them with stored high-resolution images to determine changes and freshness scores, and either order new high-resolution images or hallucinate synthetic images based on these assessments, while vehicles capture street-level images and decide on upload based on change detection and freshness scores.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If high-resolution aerial images are acquired frequently to maintain maximum coverage, then image freshness and quality are improved, but acquisition cost increases significantly
Solution Approach 1:
The patent uses low-resolution aerial images as a cheap, frequently-updatable alternative to expensive high-resolution images. These low-cost images serve as temporary placeholders that can be rapidly acquired and used to detect changes, replacing the need for frequent expensive high-resolution acquisitions while maintaining image freshness through the change detection mechanism
Solution Approach 2:
The system performs preliminary change detection using low-resolution images before committing to expensive high-resolution re-acquisition. By pre-screening areas for changes using inexpensive data, the system avoids unnecessary costly acquisitions and only orders new high-resolution images when actual changes are detected, optimizing the balance between freshness and cost
2Loss of information
If vehicles store all captured street-level images locally, then data availability is improved, but storage device capacity is quickly exhausted
Solution Approach 1:
The patent extracts only the essential information (change detection results) from the captured street-level images rather than storing the complete image datasets. The vehicle computes change metrics between captured images and previously stored images, then transmits only the necessary update information to remote computing devices, dramatically reducing local storage requirements while maintaining data availability
Solution Approach 2:
The system segments the image data processing by separating change detection computations performed locally in the vehicle from the actual image storage and archiving performed remotely. This division allows the vehicle to maintain data availability through local processing while offloading bulk storage to remote servers with unlimited capacity
3Loss of information
If all captured street-level image data is uploaded to remote computing devices, then data completeness is improved, but transmission time and bandwidth consumption increase
Solution Approach 1:
The patent applies partial action by uploading only the subset of street-level images that contain detected changes rather than uploading all captured images. The system computes change metrics and selectively transmits only those images where changes exceed a threshold, significantly reducing upload time and bandwidth consumption while maintaining data completeness for areas that actually changed
Data Source
AI summary
Methods and apparatuses are disclosed. Previously stored images of one or more geographic areas may be viewed by online users. A new low-resolution image may be acquired and aspects of the new low-resolution image may be compared with a corresponding one of the previously stored images to determine an amount of change. A determination may be made regarding whether to acquire a new high-resolution image based on the determined amount of change and a freshness score associated with the one of the previously stored images. In another embodiment, a new image may be captured and corresponding location data may be obtained. A corresponding previously stored image may be obtained and compared with the new image to determine an amount of change. The new image may be uploaded to a remote computing device based on the determined amount of change and a freshness score of the previously stored image.