Neural Network Feature Extraction for Automatic Topology Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
As the number of images increases, the cost and time required to identify and specify the positional relation between images also increase exponentially, making it inefficient to determine which images can be mapped and connected, especially when their positions are unknown.
Innovation Solution
An automatic topology mapping processing method using a neural network to extract features from images, construct a database of feature vectors, and perform vector searches to quickly identify mapping images and determine their relative positional relations, reducing the need for manual feature point extraction and comparison.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional feature point extraction and comparison methods are used to identify mapping images, then mapping accuracy can be achieved, but the time and computational resources required increase exponentially as the number of images increases
Solution Approach 1:
The patent segments the image matching process into two distinct phases: (1) feature extraction phase where deep neural networks extract comprehensive features from all images once, and (2) vector search phase where these features are quickly compared. This segmentation allows expensive feature extraction to be performed only once per image rather than repeatedly for each image pair, thereby reducing exponential computational complexity while maintaining mapping accuracy.
Solution Approach 2:
The patent performs preliminary feature extraction using deep neural networks before the actual image matching task. By pre-extracting and storing feature vectors from all images in advance, the system eliminates the need for repeated feature extraction during matching operations. This preliminary action transforms the exponentially complex O(n²) feature comparison task into a more efficient vector search operation.
2Reliability
If manual feature point extraction and pairwise image comparison is performed, then accurate mapping can be determined, but the complexity and resource consumption increase significantly with more images
Solution Approach 1:
The patent replaces the mechanical manual process of feature point extraction and pairwise comparison with an automated deep learning system. Deep neural networks automatically extract features and generate feature vectors, substituting the manual mechanical process with an intelligent automated system that reduces processing complexity while maintaining or improving mapping determination accuracy.
Solution Approach 2:
The patent changes the parameter representation from traditional hand-crafted feature points to deep learning-based feature vectors. This parameter transformation enables the use of efficient vector search algorithms instead of complex pairwise feature point matching, thereby reducing processing complexity while preserving the ability to accurately determine image mappings.
3Loss of information
If exhaustive pairwise comparison of all images is performed to find mapping relationships, then complete mapping information can be obtained, but the computational cost increases exponentially
Solution Approach 1:
The patent segments the exhaustive comparison task into feature extraction (performed once per image) and vector search (performed efficiently for each query image). This segmentation ensures that mapping information is obtained for all images without requiring O(n²) pairwise comparisons, thereby reducing computational energy consumption while maintaining information completeness.
Solution Approach 2:
The patent creates feature vector copies of all images and stores them in a database for efficient retrieval. Instead of comparing original images repeatedly, the system works with compressed feature vector representations, reducing computational energy while ensuring that mapping relationships are identified for all images through the vector search mechanism.
Data Source
AI summary
An automatic topology mapping processing method and system. The automatic topology mapping processing method includes the steps of: obtaining, by the automatic topology mapping processing system, a plurality of images, wherein at least two of the plurality of images include a common area in which a common space is captured; extracting, by the automatic topology mapping processing system, from respective images, features of the respective images through a feature extractor using a neural network; and determining, by the automatic topology mapping processing system, mapping images of the respective images on the basis of the features extracted from the respective images.