Road Profile Fusion Using Image and Point-Cloud 3D Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous driving systems face challenges in accurately determining road profiles, especially on uneven terrain, as existing methods using 3D-LaneNet may not effectively represent lanes and road features.
Innovation Solution
The system extracts image features, generates segmentation masks, and combines image-based and point-cloud-based three-dimensional features using a volumetric voxel-attention transformer to generate a road profile and perturbation map, which includes a polynomial representation of the road surface and deviations like potholes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If 3D-LaneNet is used to determine road profiles, then the system can process images to identify lanes, but it fails to accurately represent lanes and road features on uneven terrain
Solution Approach 1:
The patent combines image-based three-dimensional features with point-cloud-based three-dimensional features to create a more accurate and reliable road profile representation. This fusion of multiple data sources (images and point clouds) allows the system to overcome the limitations of using either method alone, particularly on uneven terrain where single-method approaches fail to accurately represent road features
2Device complexity
If the system uses only image data to determine road profiles, then processing is simpler, but the system cannot effectively detect road perturbations like potholes on uneven terrain
Solution Approach 1:
The patent introduces point-cloud data as an intermediary element that bridges the gap between simple image processing and accurate road perturbation detection. The point-cloud-based three-dimensional features serve as a mediator that enhances the detection capability for road perturbations like potholes, while the fusion architecture manages complexity through structured feature combination
3Loss of information
If the system processes all image features, then comprehensive road information is obtained, but processing efficiency decreases and irrelevant features interfere with road profile determination
Solution Approach 1:
The patent extracts and selects only the relevant three-dimensional features from the complete set of image features using segmentation masks. This extraction process removes irrelevant features while preserving the essential road information needed for accurate profile determination, thereby improving processing efficiency without sacrificing information completeness
Data Source
AI summary
Systems and techniques are described herein for determining road profiles. For instance, a method for determining a road profiles is provided. The method may include extracting image features from one or more images of an environment, wherein the environment includes a road; generating a segmentation mask based on the image features; determining a subset of the image features based on the segmentation mask; generating image-based three-dimensional features based on the subset of the image features; obtaining point-cloud-based three-dimensional features derived from a point cloud representative of the environment; combining the image-based three-dimensional features and the point-cloud-based three-dimensional features to generate combined three-dimensional features; and generating a road profile based on the combined three-dimensional features.