Ground-to-Aerial Image Matching via Partial Region Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing ground-to-aerial cross-view matching systems face accuracy issues due to situations where a section of the aerial-view image may include objects not present in the ground-view image, or vice versa, affecting the matching process.
Innovation Solution
An image matching apparatus that acquires and extracts features from both ground-view and aerial-view images, further extracts partial aerial regions from the aerial-view image, computes combined aerial features by combining these features, and determines matching by comparing them with the ground-view image features, considering both local and global features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the system compares features from the entire aerial-view image with ground-view image features, then the matching process is simple, but accuracy decreases when objects are not included in both views
Solution Approach 1:
The patent divides the aerial-view image into multiple partial aerial regions and compares each region separately with the ground-view image. This segmentation allows the system to focus on specific areas where matching is more reliable, improving accuracy when objects are not present in both views by eliminating irrelevant regions from the comparison.
2Productivity
If the system uses only global aerial features for matching, then computation is faster, but reliability decreases when local objects are mismatched
Solution Approach 1:
The patent transitions from a single global comparison approach to a multi-dimensional approach by evaluating multiple partial aerial regions. Each region provides an additional dimension of comparison, allowing the system to achieve more reliable matching results by considering local variations while maintaining computational efficiency through selective region processing.
Data Source
AI summary
The image matching apparatus (2000) acquires a ground-view image (20) and an aerial-view image (30). The image matching apparatus (2000) extracts features from the ground-view image (20). The image matching apparatus (2000) extracts features from the aerial-view image (30). The image matching apparatus (2000) extracts a plurality of partial aerial regions (32) from the aerial-view image (30), and extracts features from each partial aerial region (32). The image matching apparatus (2000) computes, for each partial aerial region (32), a combined aerial feature by combining the features of the partial aerial region (32) and the features of the aerial-view image (30). The image matching apparatus (2000) determines, for each partial aerial region (32), whether the partial aerial region (32) matches the ground-view image (20) by comparing the combined aerial feature of the partial aerial region (32) and the features of the ground-view image (20).