Moving Object Localization via 2D-3D Landmark Probability Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous driving systems face challenges in accurately estimating the position of moving objects, particularly due to the dimension mismatch between 2D feature point information from surrounding images and 3D feature point information from high-definition maps, which complicates similarity calculation and position estimation.
Innovation Solution
The method involves generating 2D feature point information from surrounding images and 3D feature point information from high-definition maps, converting these into compatible dimensions using perspective mapping, and calculating similarity by summing and multiplying probabilities corresponding to landmarks, with the processor implementing particle filter or maximum likelihood optimization to estimate the object's position.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If 2D feature point information from surrounding images and 3D feature point information from HD maps are used directly for position estimation, then the system can utilize both real-time visual data and pre-defined map data, but the dimension mismatch between 2D and 3D feature points complicates similarity calculation and reduces estimation accuracy
Solution Approach 1:
The patent converts 3D feature point information from HD maps into 2D coordinates using perspective mapping, transforming the dimensionality to match the 2D feature points from surrounding images. This dimensional transformation enables direct comparison and similarity calculation between map-based and image-based feature points, resolving the dimension mismatch problem while maintaining position estimation accuracy
2Measurement precision
If dimension conversion between 2D and 3D feature points is performed to enable similarity calculation, then position estimation accuracy can be improved, but additional processing steps and computational complexity are introduced
Solution Approach 1:
The patent performs perspective mapping conversion of 3D feature points to 2D coordinates as a preliminary step before similarity calculation. By pre-converting the dimensional format of map feature points, the system eliminates the need for complex real-time dimensional transformation during position estimation, reducing processing time while maintaining accuracy
Data Source
AI summary
To estimate a position of a moving object, an electronic device generate two-dimensional (2D) feature point information of a landmark-based probability map from the surrounding image, obtain landmark-based three-dimensional (3D) feature point information from the HD map data, convert one of the 2D feature point information of the surrounding image to 3D or the 3D feature point information of the HD map data to 2D, determine a similarity between the converted feature point information and one of the feature point information of the surrounding image and the feature point information of the HD map data that is not converted, and estimate a position of the moving object based on the similarity.