Lane Map Object Localization Using Correlated Image Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining the location of objects on a lane map in autonomous driving systems are inefficient, as they require processing large amounts of image data to identify and track objects like traffic lights and road signs, which increases computational load and reduces real-time processing capabilities.
Innovation Solution
A method and apparatus that acquire image data from a vehicle-mounted camera, generate feature information by processing the data, and determine the location of objects on the lane map by correlating features across multiple images, using metadata to identify and track objects without relying on entire image data sets, thereby reducing computational requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If entire image data is processed to identify and track objects, then object location determination accuracy is improved, but computational load increases and real-time processing capability deteriorates
Solution Approach 1:
The patent extracts only the necessary feature information (metadata) from image data for object identification and tracking, rather than processing entire images. This extraction approach maintains object location accuracy while significantly reducing computational load and enabling real-time processing.
Solution Approach 2:
The patent segments the object identification process into two stages: first extracting feature information from images, then using this feature data for tracking and location determination. This segmentation allows processing of only essential data elements, improving real-time capability while maintaining accuracy.
2Productivity
If feature information is extracted and correlated across multiple images, then data processing load is reduced, but object tracking accuracy may deteriorate
Solution Approach 1:
The patent implements feedback mechanisms in the object tracking process, where feature information from multiple images is correlated and compared. This feedback loop ensures that object tracking maintains accuracy by continuously verifying object identity across frames while processing only extracted feature data rather than full images.
Data Source
AI summary
Provided are a method and apparatus for determining a location of an object on a lane map, wherein the method includes acquiring image data captured by a camera mounted on a vehicle, generating feature information regarding a plurality of features included in the image data by performing certain processing on the image data, determining a plurality of features corresponding to a same object in a plurality of pieces of image data, on the basis of a degree of correlation between the plurality of features included in the image data, and determining a location of an object on the lane map, on the basis of location information of the vehicle on the lane map and feature location information on the plurality of pieces of image data.