HD Map Construction via Hybrid Point-Element Queries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing HD map construction methods rely solely on point query representation, which struggles to accurately represent the details of map elements, leading to incomplete shapes and inaccurate positions in the constructed maps.
Innovation Solution
The proposed method employs a hybrid query combining point and element queries to extract and integrate point-level and element-level information, using a hybrid decoder to iteratively update map information and construct a high-definition map with enhanced accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If point query representation is used for HD map construction, then the method is simple, but the map element details are incomplete and positions are inaccurate
Solution Approach 1:
The query representation is segmented into two distinct components: point queries for location information and element queries for shape and semantic information. This segmentation allows each component to specialize in extracting specific types of features, thereby improving overall map element accuracy while maintaining a manageable system structure.
Solution Approach 2:
The method transitions from a single-dimension point query representation to a two-dimension hybrid representation by introducing element queries as an additional dimension. This dimensional expansion enables the system to capture both point-level location details and element-level shape characteristics simultaneously, resolving the accuracy limitation.
2Productivity
If point query representation is used for HD map construction, then the processing is fast, but the map element shapes are inaccurate
Solution Approach 1:
The processing task is segmented into two parallel streams: point query processing for location detection and element query processing for shape reconstruction. This segmentation allows each stream to be optimized independently, maintaining processing efficiency while improving shape accuracy through specialized element-level feature extraction.
Solution Approach 2:
The method merges the outputs of point queries and element queries in the hybrid decoder to produce the final HD map. This merging combines the speed advantages of point-based detection with the accuracy benefits of element-based shape representation, achieving both fast processing and accurate shape reconstruction.
3Use of energy by moving object
If point query representation is used for HD map construction, then the computational load is low, but the map element details are incomplete
Solution Approach 1:
The information extraction process is segmented into point-level feature extraction and element-level feature extraction. This segmentation enables the system to capture detailed map element information that would be lost in a purely point-based approach, while distributing the computational load across two specialized processing paths.
Solution Approach 2:
By adding the element query dimension to the representation, the system recovers lost map element detail information without proportionally increasing computational load. The element queries are designed to efficiently capture shape and semantic features that complement the point query information.
Data Source
AI summary
A high-definition (HD) map-related map construction method, electronic device, and storage medium are provided. The method includes: extracting a bird's-eye view (BEV) feature map based on the data; determining map information through a hybrid decoder based on the BEV feature map and a hybrid query; and constructing an HD map corresponding to the data based on the map information, wherein the map includes a plurality of map elements each including an area formed by a plurality of coordinate points in the map, the map information comprises coordinate information and class information of the plurality of map elements, and the hybrid query includes a plurality of hybrid features each corresponding to one map element and including a point feature and an element feature. Optionally, the method may be executed using an artificial intelligence (AI) model.