Graph Convolutional Map Fusion via Neural Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current map fusion techniques are time-consuming, resource-intensive, and prone to inaccuracies due to their dependency on parameter tuning and inability to adapt to changing street patterns, often requiring human intervention and suffering from high runtime complexity.
Innovation Solution
A trainable map fusion method using graph convolutional neural networks that processes source and target maps through graph convolutional layers, linear rectifying layers, and fully connected layers to produce node feature maps and probability matching outputs, enabling efficient and accurate fusion of digital street maps.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional map fusion techniques are used to merge multiple digital street maps, then comprehensive map information can be obtained, but the process becomes time-consuming and resource-intensive with high runtime complexity
Solution Approach 1:
The patent replaces traditional mechanical/map fusion algorithms with a deep learning-based neural network system. The system uses neural networks to automatically learn and extract features from multiple map sources, substituting complex computational algorithms with trained models that achieve faster processing while maintaining or improving fusion accuracy.
Solution Approach 2:
The patent performs preliminary processing of map data before fusion by pre-training neural network models on map features. This preliminary action includes pre-extracting features from individual maps and preparing data structures in advance, which significantly reduces the computational burden during actual fusion operations and decreases processing time.
2Reliability
If traditional map fusion techniques are used to handle large amounts of map information, then comprehensive coverage is achieved, but computational resources are excessively consumed
Solution Approach 1:
The patent extracts only the most relevant and discriminative features from complete map data using neural network-based feature extraction. Instead of processing all map information, the system identifies and extracts key features such as road intersections, street names, and geometric characteristics, significantly reducing computational resource requirements while preserving fusion accuracy.
Solution Approach 2:
The patent segments the map fusion process into distinct computational stages handled by different neural network components. The system divides feature extraction, feature matching, and fusion decision-making into separate processing modules, allowing efficient resource allocation and parallel processing that reduces overall computational burden.
3Measurement precision
If parameter tuning is used in map fusion techniques to improve accuracy, then fusion precision can be enhanced, but the process requires human intervention and becomes more complex
Solution Approach 1:
The patent implements self-service through automatic hyperparameter optimization using neural networks. The system automatically adjusts and optimizes matching parameters and thresholds based on training data without requiring manual human intervention. This self-configuring capability maintains high matching precision while eliminating the complexity associated with manual parameter tuning.
Solution Approach 2:
The patent transforms fixed, manually-tuned parameters into dynamic, learnable parameters within the neural network model. The system automatically adapts parameter values during training based on the characteristics of different map sources, enabling precise matching while reducing system complexity by eliminating manual parameter configuration.
4Reliability
If conventional map fusion methods are used to handle diverse map formats and attributes, then all map information can be integrated, but the methods lack adaptability to changing street patterns
Solution Approach 1:
The patent implements dynamic adaptability through neural networks that can automatically adjust to different map formats, attributes, and street patterns. The system uses dynamic feature extraction that adapts to varying input characteristics and employs learned matching criteria that flexibly handle diverse map representations and evolving street configurations without requiring method redesign.
Data Source
AI summary
A map fusing method includes receiving a source graph and a target graph. The source graph is representative of a source map and the target graph is representative of a target map and includes nodes and edges that connect the nodes. The method further includes processing each of the source graph and the target graph in a graph convolutional layer to provide graph convolutional layer outputs related to the source graph and to the target graph, processing each of the graph convolutional layer outputs for the source graph and the target graph in a linear rectifying layer to output node feature maps related to the source graph and the target graph. The method further includes selecting pairs of node representations from the node feature maps related to the source graph and the target graph and concatenating the selected pairs to output selected and concatenated pairs of node representations.