Image Feature Matching With Dense Descriptor Maps for HD Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing high-definition map technologies for autonomous vehicles lack efficient methods for processing images and training multi-task processing models, which are crucial for precise positioning, obstacle detection, and collision avoidance, leading to potential safety risks.
Innovation Solution
A method of processing images to extract feature points and dense descriptor maps, determining matched feature and pixel pairs, and training a multi-task processing model using labeled samples to improve timeliness and accuracy in tasks such as visual odometry and three-dimensional reconstruction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional image processing methods are used for autonomous vehicle positioning and navigation, then the system structure is simple, but the positioning accuracy and obstacle detection precision are insufficient
Solution Approach 1:
The patent segments the image processing task into multiple specialized modules: feature point extraction module, descriptor map generation module, feature point matching module, and pixel-level matching module. Each module focuses on a specific aspect of the processing pipeline, enabling higher precision through specialized processing while managing complexity through modular design.
Solution Approach 2:
The patent transitions from traditional sparse feature point matching to dense descriptor map matching, adding a dimension of detail by processing pixel-level information. This dimensional enhancement enables much higher positioning accuracy and obstacle detection precision by utilizing comprehensive image data rather than limited feature points.
2Adaptability or versatility
If multi-task processing models are implemented for simultaneous positioning and obstacle detection, then the comprehensiveness of navigation functions is improved, but the processing time and computational load increase
Solution Approach 1:
The patent merges multiple navigation functions (positioning, obstacle detection, path recognition) into a unified multi-task processing model that processes images through shared feature extraction and descriptor generation stages. This combining approach enables comprehensive navigation functions while reducing redundant processing compared to separate specialized systems.
Solution Approach 2:
The patent performs preliminary feature extraction and descriptor map generation once for the entire image, then reuses these intermediate results across multiple tasks including positioning, obstacle detection, and path recognition. This preliminary processing avoids redundant computations and reduces overall processing time despite the comprehensiveness of multiple navigation functions.
3Measurement precision
If feature point matching is used for image alignment, then the computational complexity is low, but the matching accuracy and detail precision are insufficient
Solution Approach 1:
The patent introduces descriptor maps as an intermediary representation between feature point detection and final matching. The descriptor maps encode detailed local image characteristics around each feature point, enabling more accurate matching by comparing comprehensive descriptors rather than relying solely on feature point coordinates and basic similarity metrics.
Data Source
AI summary
A method of processing an image, a method of training a multi-task processing model, and an electronic device, which relate to a field of an automatic driving technology, in particular to a field of high-definition map technology. The method of processing an image includes: processing a to-be-processed image to obtain a feature point of the to-be-processed image, a feature point descriptor map of the to-be-processed image, and a dense descriptor map of the to-be-processed image; determining a pair of matched feature points between the to-be-processed image and a reference image based on the feature point and the feature point descriptor map; and determining a pair of matched pixels between the to-be-processed image and the reference image based on the dense descriptor map.