Map Geometry Generation Using Probe Data and Image Categorization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current digital map generation methods relying on machine learning face challenges in areas with poor georeferenced road geometry coverage, leading to inaccurate map representations, especially in developing countries where road network data is inadequate.
Innovation Solution
A method that utilizes received images and probe data to categorize pixels, determining target centers and confidence values to generate map geometry, allowing for the creation of accurate map representations even in areas with insufficient georeferenced data by associating probe data density and direction of travel.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If machine learning is used to identify target objects in images, then map generation can be automated, but map accuracy deteriorates in areas with poor georeferenced road geometry coverage
Solution Approach 1:
The patent introduces probe data as an intermediary element that bridges the gap between automated machine learning processing and accurate map representation. Probe data from mobile devices provides ground truth information about road locations and characteristics, which then guides the machine learning model to achieve both automation and accuracy in areas with poor existing georeferenced data.
Solution Approach 2:
The system incorporates feedback mechanisms where probe data from mobile devices is used to validate and correct machine learning predictions. The feedback loop allows the system to iteratively improve its accuracy by comparing automated detections with actual probe data, enabling continuous refinement of map representations in data-scarce regions.
2Ease of manufacture
If georeferenced road geometry is used to seed machine learning, then map generation can proceed with existing data, but map coverage deteriorates in developing countries with poor road network data
Solution Approach 1:
The patent applies preliminary action by collecting and preparing probe data from mobile devices before the main map generation process. This preliminary data collection in areas with poor coverage provides the necessary seed information for machine learning models to accurately generate map representations in developing countries, expanding coverage without requiring extensive pre-existing georeferenced road geometry.
3Measurement precision
If probe data is used to categorize pixels, then map precision can be improved in data-scarce regions, but processing complexity increases
Solution Approach 1:
The patent segments the complex task of map generation into distinct stages: probe data collection, pixel categorization based on probe data density, and map geometry generation. This segmentation allows the system to handle complexity systematically by processing data in manageable chunks and using probe data to guide subsequent processing steps, thereby improving precision without overwhelming computational complexity.
Data Source
AI summary
A method, apparatus and computer program product are provided for generating map geometry based on a received image and probe data. A method is provided including receiving a first image and probe data associated with the first image, categorizing pixels of the first image based on the probe data, and generating a map geometry based on the pixel categorization of the first image.