Aerial Image Feature Extraction Using Terrestrial Control Points
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Creating high-accuracy maps is costly and challenging due to the reliance on positioning technology, which loses accuracy in urban environments, and manual extraction of map features from images is inefficient and scalable.
Innovation Solution
A method involving a learned model based on control points collected from terrestrial data, where aerial images are orthorectified and used to identify mapping information for a map database, utilizing an image correlation module, learned model training device, and inference module to automate feature extraction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If industrial capture vehicles with high-quality devices are deployed to achieve high-accuracy maps, then mapping precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent replaces complex mechanical positioning systems (industrial capture vehicles with high-quality sensors) with a learned model-based image processing system. The system uses aerial images combined with readily available terrestrial data (GPS coordinates, map data) to train a learned model that automatically extracts road network features, substituting the need for expensive mechanical capture vehicles while achieving comparable or superior mapping precision.
Solution Approach 2:
The patent creates a learned model that copies the mapping functionality from training data (aerial images with associated terrestrial coordinates) to new unseen areas. Instead of deploying physical capture vehicles to every location, the system learns from example data and applies the learned patterns to generate accurate maps of new regions, significantly reducing device complexity and deployment cost.
2Manufacturing precision
If manual extraction of map features from images is performed, then mapping precision is improved, but productivity decreases
Solution Approach 1:
The patent implements a self-service system where the learned model automatically extracts road network features from aerial images without human intervention. The system trains on labeled training data and then autonomously processes new images to identify roads, intersections, and other map features, eliminating the need for manual feature extraction while maintaining high accuracy and dramatically improving productivity.
Solution Approach 2:
The patent introduces a learned model as an intermediary between raw aerial images and final map features. This intermediary system automatically performs the feature extraction task that would otherwise require manual human analysis, serving as a bridge that translates image data into structured map information with high efficiency and consistency.
3Ease of operation
If positioning technology is used in urban environments, then vehicle position is obtained, but measurement precision deteriorates
Solution Approach 1:
The patent uses feedback from multiple data sources (aerial images, terrestrial GPS data, existing map data) to correct and improve positioning accuracy. The learned model compares detected features in aerial images with known features from training data and map databases, using this feedback to refine position estimates and compensate for positioning technology errors in urban environments where satellite signals are blocked or reflected.
Data Source
AI summary
An apparatus, or corresponding method, for building or updating a map database is described. In one example, the apparatus includes an image correlation module, a training device, and a learned model or neural network. The image correlation module is configured to correlate a first aerial image and terrestrial sensor data collected at a terrestrial vehicle based on at least one control point from the terrestrial data. The learned model training device is configured to define a learned model based using at least one control point from the terrestrial sensor data as ground truth for analysis of the first aerial image. The learned model inference module is configured to receive a second aerial image and apply the learned model on the second aerial image for identification of mapping information for the map data.