3D Feature Point Selection for Balanced Vehicle Map Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional map generation apparatuses for self-driving vehicles face accuracy issues due to uneven distribution of feature points near and far from the camera, leading to inaccurate estimation of vehicle position and posture.
Innovation Solution
A vehicle control apparatus that extracts feature points from detection information, selects points to balance their distribution across different distance ranges, and calculates three-dimensional positions to generate accurate maps, reducing bias and improving map accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If feature points are extracted from captured images without selective filtering, then the quantity of feature points is increased, but the distribution bias between near and far objects causes impaired map accuracy
Solution Approach 1:
The patent applies local quality by differentiating the treatment of feature points based on their spatial characteristics. Feature points are categorized into near objects and far objects, and selective filtering is applied differently to each category. This ensures that the quality of map information is optimized for each local region rather than applying a uniform approach to all feature points.
Solution Approach 2:
The patent changes the parameter of feature point selection by introducing distance-based filtering criteria. By adjusting the selection parameters according to the distance of objects from the vehicle, the system optimizes the contribution of feature points at different ranges, thereby improving overall map accuracy while maintaining an adequate quantity of feature points.
2Productivity
If all extracted feature points are used for map generation, then processing efficiency is maintained, but the bias in feature point distribution between distance ranges degrades estimation accuracy of vehicle position and posture
Solution Approach 1:
The patent applies preliminary action by performing selective filtering of feature points before the map generation process. By pre-processing the feature points to remove biased samples from near objects while preserving those from far objects, the system ensures that the subsequent map generation and vehicle estimation processes work with optimized data, improving accuracy without significantly impacting efficiency.
3Quantity of substance
If feature points from near objects are heavily weighted, then the quantity of available feature points is increased, but the map information becomes biased and less accurate for distant objects
Solution Approach 1:
The patent applies the anti-weight principle by counterbalancing the influence of feature points from near objects with those from far objects. Through selective filtering, the system reduces the excessive weight of near object feature points while maintaining or enhancing the contribution of far object feature points, thereby achieving a balanced and reliable map representation that accurately reflects the entire environment.
Data Source
AI summary
A vehicle control apparatus includes a microprocessor configured to perform: extracting feature points from detection information detected by an in-vehicle detection unit; selecting feature points for which three-dimensional positions are to be calculated from extracted feature points; based on a plurality of detection information, calculating three-dimensional positions of same feature points in the plurality of detection information for the selected feature points using a position and posture of the in-vehicle detection unit; and generating a map including information of each of the three-dimensional positions using the calculated three-dimensional positions of the plurality of the feature points. The selecting includes selecting the feature points so as to reduce a bias in a number of feature points on objects located in a first distance range unit and a number of feature points on objects located in a second distance range farther than the first distance range.