Self-position estimation using depth variation weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing self-position estimation systems for moving objects in automatic driving and driver assistance face challenges in accurately determining the position of objects with varying observation points, such as plants and inclined banks, due to changes in depth observation when the object's position changes slightly, leading to low accuracy in position estimation.
Innovation Solution
A self-position estimation apparatus that calculates a correspondence weight between map data and observation data based on the variation amount of depth in a predetermined direction, using a distance sensor to scan observation vectors and determine valid areas for alignment, thereby improving the accuracy of self-position estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If depth observation is used to determine object positions for self-position estimation, then position determination can be performed using map data and observation data, but accuracy deteriorates when objects have varying observation points (such as plants and inclined banks) because depth values change significantly with slight position changes
Solution Approach 1:
The patent applies local quality by assigning different weights to different spatial regions based on their depth variation characteristics. Areas with high depth variation (like plants and inclined banks) are identified and given lower weights or excluded from alignment, while stable areas maintain higher weights. This localized differentiation allows the system to handle varying observation points appropriately without compromising overall estimation accuracy.
Solution Approach 2:
The patent changes the parameter of depth observation by introducing depth variation amount as a new criterion. Instead of using absolute depth values for alignment, the system calculates how much depth changes in specific directions and uses this variation information to determine valid alignment areas. This parameter transformation converts the problem from direct depth matching to variation-based filtering.
2Measurement precision
If all observation data is used for alignment to improve estimation accuracy, then more information is utilized, but errors increase when objects with varying observation points are included
Solution Approach 1:
The patent segments the observation data into valid and invalid areas based on depth variation analysis. By dividing the spatial domain and identifying which regions exhibit excessive depth variation, the system selectively processes only reliable data segments for alignment. This segmentation prevents problematic objects from contaminating the overall estimation while preserving useful information from stable regions.
Solution Approach 2:
The patent extracts and removes observation data from areas with high depth variation before performing alignment. By identifying regions where depth changes significantly in the depth direction and excluding these from the alignment process, the system eliminates harmful information while retaining beneficial data from stable areas, thereby improving overall estimation reliability.
Data Source
AI summary
A self position estimation apparatus according to an embodiment includes an acquisition unit configured to acquire map data, and a calculation unit configured to calculate a correspondence weight between observation data on a region around the moving object and the map data, based on a variation amount of depth in a predetermined direction.