3D Road Surface Reconstruction with Horizontal-Line ROI Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for 3D road surface reconstruction in autonomous driving suffer from significant noise in road surface estimation, leading to inaccuracies in lane marking detection and increased errors in 3D road surface reconstruction.
Innovation Solution
A method and apparatus that combine horizontal line detection and lane marking detection to improve road surface estimation and compensation during 3D road surface reconstruction, by calibrating horizontal line information and adjusting Region of Interest (ROI) information based on compensation values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional computer vision methods are used for road surface estimation, then lane marking detection can be performed, but considerable noise is introduced leading to inaccurate 3D road surface reconstruction
Solution Approach 1:
The patent introduces horizontal line information as an intermediary element to mediate between the image data and the 3D road surface reconstruction process. By detecting horizontal lines in the image and using their pixel values as reference, the system creates a stable intermediary reference that reduces noise in the road surface estimation, thereby improving both measurement precision and reliability of lane marking detection
Solution Approach 2:
The patent implements a feedback mechanism where the horizontal line pixel values are used to compensate and adjust the ROI information during 3D road surface reconstruction. The system continuously monitors the horizontal line positions and uses this feedback to correct deviations in the reconstructed road surface, thereby maintaining accurate lane marking detection despite variations in road conditions
2Measurement precision
If Hough transform and least squares method are applied for 3D estimation, then road surface reconstruction can be performed, but a large amount of convolutional computation is required making it inefficient for embedded systems
Solution Approach 1:
The patent extracts only the essential horizontal line information from the image rather than performing comprehensive edge detection and convolutional computations. By taking out just the horizontal line pixel values and positions, the system achieves 3D road surface estimation without the heavy computational burden of Hough transform and least squares methods, thereby improving productivity while maintaining measurement precision
Solution Approach 2:
The patent segments the complex road surface estimation problem into simpler components: first detecting horizontal lines, then using their pixel values to directly compute road surface parameters. This segmentation avoids the need for full-image convolutional operations and complex fitting procedures, making the system efficient for embedded systems while preserving estimation accuracy
Data Source
AI summary
The present disclosure provides a method, apparatus and computer program product for suppressing noise in 3D road surface reconstruction. The method includes: acquiring an image related to a road surface; extracting horizontal line information and ROI (region of interest) information from the image, wherein the horizontal line information comprises a horizontal line pixel value, and the ROI information is used to perform 3D road surface reconstruction for the road surface; judging whether the difference between the horizontal line pixel value and a preset horizontal line pixel value is greater than a pixel threshold; judging whether the horizontal line pixel value is greater than a compensation threshold when the difference is greater than the pixel threshold; determining a compensation value based on the difference when the horizontal line pixel value is not greater than the compensation threshold; and adjusting the ROI information based on the compensation value.