Image Matching via Multi-Feature Vectors for Accurate Map Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image matching methods using convolutional neural networks for map updates suffer from inaccurate element recognition due to limited feature extraction, leading to high error rates.
Innovation Solution
Perform feature extraction on images to obtain feature maps, determine feature vectors for each point, and match these vectors to improve understanding and accuracy of image similarity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If target detection network is used for element recognition, then image processing can be automated, but element recognition accuracy deteriorates due to limited feature extraction
Solution Approach 1:
The patent segments the feature extraction process into multiple independent feature maps, each capturing different aspects of image features. Instead of relying on a single target detection network output, the system divides feature extraction into multiple channels that can be independently processed and then combined, thereby improving overall recognition accuracy while maintaining automation.
Solution Approach 2:
The patent combines multiple types of feature maps (different feature extraction results) into a composite feature representation. By fusing multiple feature sources rather than relying on a single detection network output, the system creates a more comprehensive and accurate element recognition capability that maintains full automation.
2Measurement precision
If multiple feature maps are extracted and combined into feature vectors, then image matching accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent extracts only the essential and most discriminative features from multiple feature maps to construct feature vectors. Rather than processing all possible features equally, the system selectively extracts key features that most contribute to matching accuracy, thereby reducing computational complexity while maintaining or improving matching performance.
Solution Approach 2:
The patent applies partial action by selecting and processing only the most relevant feature maps and feature points necessary for accurate matching. Instead of exhaustively processing all features from all feature maps, the system identifies and processes the critical subset that provides the most value, balancing accuracy with computational efficiency.
Data Source
AI summary
An image matching method includes: performing feature extraction processing on a first image to obtain K first feature maps; performing feature extraction processing on a second image to obtain K second feature maps; determining a first feature vector of each of M first feature points in the first image based on the K first feature maps, to obtain M first feature vectors; determining a second feature vector of each of N second feature points in the second image based on the K second feature maps, to obtain N second feature vectors; determining a quantity of feature point pairs based on the M first feature vectors and the N second feature vectors; and determining an image matching result based on the quantity of feature point pairs.