HD Map Correction via Crowd-Sourced Image Stitching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current navigation maps, such as OpenStreetMap and Google Maps, are not suitable for autonomous vehicles as they lack the precision and detail required for high-definition mapping, which can lead to errors due to environmental changes and outdated data.
Innovation Solution
A system utilizing crowd-sensing technology to identify map anomaly hotspots by synthesizing sequential ground view images from a fleet of vehicles, converting them to bird view images, removing outliers, applying error reduction algorithms, and comparing them to existing high-definition maps to provide revised imagery for autonomous vehicles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If navigation maps are updated frequently to reflect environmental changes, then map accuracy is improved, but cost and time consumption increase
Solution Approach 1:
The system enables maps to self-update by automatically detecting anomalies through sensor data from vehicles and crowd-sourced images, eliminating the need for manual validation and frequent scheduled updates. The anomaly detection system continuously monitors the environment and triggers updates only when changes are detected, allowing the mapping system to service itself.
Solution Approach 2:
The system changes the update frequency parameter dynamically based on detected anomalies rather than using fixed intervals. When map anomalies are detected through sensor data or crowd-sourced images, the system triggers targeted updates at those specific locations and times, optimizing both accuracy and resource utilization.
2Reliability
If manual validation is performed to ensure map accuracy, then map reliability is improved, but cost increases
Solution Approach 1:
The system replaces manual validation mechanics with automated sensor-based detection and image processing algorithms. Vehicles equipped with sensors automatically detect map anomalies, and computer vision algorithms validate the data, eliminating the need for human reviewers while maintaining or improving reliability through continuous automated monitoring.
Solution Approach 2:
The system implements feedback loops where sensor data and crowd-sourced images are continuously compared against existing map data, automatically triggering validation and update processes when discrepancies are detected. This closed-loop feedback system ensures high reliability through automated verification without requiring manual intervention.
3Measurement precision
If high-definition maps include every detail with precise location, then navigation precision is improved, but device complexity increases
Solution Approach 1:
The system segments the high-definition map into discrete anomaly hotspots rather than treating the entire map as a single complex entity. Each detected anomaly becomes an independent update target, allowing the system to manage complexity by dividing the large-scale mapping problem into smaller, localized correction tasks that can be processed independently.
4Productivity
If crowd-sourcing is used to collect map data, then productivity is improved, but data quality control becomes more difficult
Solution Approach 1:
The system introduces sensor data from vehicles as an intermediary layer between crowd-sourced images and the final map update. Sensors provide objective, machine-measured data that validates and filters crowd-sourced content, acting as a mediator that ensures data quality while maintaining the productivity benefits of crowd-sourcing by automatically filtering out low-quality submissions.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach provides more accurate and up-to-date high-definition mapping data, enhancing navigation precision and reducing the costs associated with manual validation and frequent updates.
Implementation Method 1
convert the ground view image data to bird view image data using homography
Data Source
AI summary
A system for correcting HD map images is provided. The system is configured to: receive, from a plurality of vehicles, ground view image data of a map anomaly hotspot captured by the vehicles; convert the ground view image data to bird view image data using homography; remove outliers from the bird view image data; apply an error reduction algorithm; stitch together a sequence of consecutive images from the image data after error reduction to produce a high-definition image, wherein each of the consecutive images has been generated through converting ground view image data to bird view image data using homography, removing outliers from the bird view image data, and applying an error reduction algorithm; compare the stitched sequence of consecutive images to a high-definition image of the hotspot to determine revisions to the high-definition image; and provide revised high-definition image data for use in navigating near the map anomaly hotspot.