Map Image Geolocation Inference via CNN and Nearest Neighbor
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods fail to effectively extract geolocation information from physically displayed or printed maps without metadata, limiting users' ability to capture and utilize this information with electronic devices for navigation.
Innovation Solution
A computer-implemented method using a convolutional neural network to generate style-invariant and location-variant map representations, combined with a nearest neighbor algorithm for image comparison, allows association of non-electronic maps with electronic maps, enabling geolocation extraction and overlay display.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a brute force approach is used to manually obtain and compare various electronic maps, then a closest match can be determined, but the process is not available in automated real-time manner
Solution Approach 1:
The system performs preliminary actions by pre-processing map images into style-invariant representations and organizing them in a database before real-time queries. This allows the automated system to quickly compare captured images against pre-processed map data without performing complex manual comparisons during real-time operation.
Solution Approach 2:
The patent introduces an intermediary processing layer that transforms captured map images into a standardized representation format suitable for automated comparison. This intermediary step enables real-time processing by converting diverse map styles into a common language that can be efficiently matched against the database.
2Loss of information
If physical maps are used for navigation, then users can orient themselves to a new location, but users cannot capture the relevant information on the map and take it with them as they physically move away
Solution Approach 1:
The system creates a digital copy of the physical map information by capturing an image of the physical map and processing it through the automated recognition system. This digital copy can then be stored and accessed electronically, allowing users to retain map information without physically carrying the original map.
Solution Approach 2:
The patent replaces the mechanical process of manually capturing and transcribing map information with an automated image processing system. The convolutional neural network automatically extracts and recognizes map data, substituting manual information capture with automated digital processing.
3Loss of information
If maps are permanently affixed to signage, then they provide orientation information, but they do not include any indicators such as metadata
Solution Approach 1:
The system introduces an intermediary processing layer that extracts metadata information from the visual content of physical maps. By analyzing the map image through automated recognition, the system infers location data and other metadata without requiring physical metadata indicators to be present on the map itself.
Solution Approach 2:
The patent replaces the physical requirement for metadata indicators with automated digital extraction. Instead of relying on printed metadata on physical maps, the system uses image processing and pattern recognition to automatically derive location information and metadata from the map content.
Data Source
AI summary
A computer implemented method of associating a non-electronic map with an electronic map is provided. The method comprises, for an obtained online data collection, generating a database of style-invariant and location-variant map representations by application of a convolutional neural network. Additionally, for a captured image associated with the non-electronic map, the captured image not having geo-location metadata, the method comprises applying a nearest neighbor heuristic to compare the captured image to the map representations and generate a match between the captured image and the map representations, and performing a display action associated with the match.