Sparse Map Image Harvesting for Autonomous Vehicle Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in navigating efficiently due to the vast amount of data they need to process and store, particularly with traditional mapping technologies, which can lead to storage and update issues.
Innovation Solution
The system employs a processor to enrich navigational maps using previously captured images, identifying objects and localizing them within a map database using object models, and selectively collects images from multiple vehicles to update the map, reducing data storage needs through a sparse map approach.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional mapping technology is used to navigate, then navigation accuracy is improved, but data storage requirements and map update complexity increase significantly
Solution Approach 1:
The patent extracts only the essential visual features and landmark representations from captured images, storing only these extracted representations rather than the complete image data. This allows the system to maintain navigation accuracy through feature-based mapping while dramatically reducing storage requirements by discarding redundant pixel-level data.
Solution Approach 2:
The patent segments the mapping process into discrete landmark detections and feature point representations, organizing spatial data into a sparse graph structure rather than storing continuous dense map data. This segmentation enables efficient storage and processing while preserving navigation-critical information.
2Measurement precision
If traditional mapping technology is used to navigate, then navigation accuracy is improved, but system complexity and processing requirements increase
Solution Approach 1:
The patent extracts and stores only the critical visual features and landmark representations from captured images, filtering out redundant information. This extraction approach simplifies the map data structure and reduces processing complexity while maintaining navigation accuracy through the preserved essential features.
Solution Approach 2:
The patent performs preliminary processing of captured images to identify and extract only the essential features and landmarks before storing them in the map database. This preliminary action reduces the complexity of subsequent navigation processing by pre-simplifying the data structure.
3Reliability
If all captured images are stored for navigation, then navigation reliability is improved, but data management and storage costs increase
Solution Approach 1:
The patent extracts only the essential visual features and landmark representations from captured images, storing these condensed representations rather than the complete image data. This maintains navigation reliability through feature-based reference points while dramatically reducing data volume by discarding redundant pixel-level information.
Solution Approach 2:
The patent creates a simplified copy of the essential features and landmarks from the original captured images, storing this reduced representation in the map database. This copying approach preserves the critical navigation information while reducing the amount of data that needs to be managed and stored.
Data Source
AI summary
Systems and methods for harvesting images for vehicle navigation are disclosed. In one implementation, a system includes a processor configured to receive an image having been captured by a camera of a vehicle from an environment of the vehicle during a first time period; during a second time period after the first time period: use a first object model to identify a representation of a first object in the image and localize the image relative to a map database based on the representation of the first object; and store the image in the map database based on the localization; and during a third time period after the second time period, use a second object model to identify a representation of a second object in the image, wherein the second object model was not available to the system during the second time period.