3D Lane Marking Detection with B-Spline Curve Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for 3D lane detection, such as 3D-LaneNet and Gen-LaneNet, require post-processing steps to convert discrete predicted points into continuous line models, and grid-based methods like 3D-LaneNet+ need clustering, resulting in discrete line representations that are less accurate and require costly post-processing.
Innovation Solution
A method using a neural network trained with B-spline functions to directly predict control points of a continuous curve for 3D lane representation, allowing for accurate and efficient detection of lane markings without the need for post-processing, by parameterizing the geometry of lane markings as a continuous curve.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If anchor-based methods or grid-based methods are used for 3D lane detection, then lane markings can be detected, but post-processing steps are required to convert discrete predicted points into continuous line models
Solution Approach 1:
The neural network is pre-trained with B-spline function parameters during the training phase, so that when detecting lane markings, it directly outputs control points of continuous curves without requiring post-processing. The preliminary action of training with B-spline parameters enables the network to inherently understand continuous curve representation
Solution Approach 2:
The invention changes the output parameters of the neural network from discrete point coordinates to B-spline control point parameters. By parameterizing the lane geometry using B-spline functions with control points, the network transforms discrete predictions into continuous curve representations through parameter transformation rather than geometric reconstruction
2Productivity
If grid-based methods with clustering are used, then lane detection can be performed, but the lines remain discrete (string of small straight lines) and require clustering post-processing
Solution Approach 1:
The neural network is pre-trained with B-spline function parameters during the training phase, so that when detecting lane markings, it directly outputs control points of continuous curves without requiring post-processing. The preliminary action of training with B-spline parameters enables the network to inherently understand continuous curve representation
Solution Approach 2:
The invention changes the output parameters of the neural network from discrete point coordinates to B-spline control point parameters. By parameterizing the lane geometry using B-spline functions with control points, the network transforms discrete predictions into continuous curve representations through parameter transformation rather than geometric reconstruction
3Device complexity
If discrete line representations are used, then computation is simpler, but accuracy is reduced due to interpolation errors and discrete nature
Solution Approach 1:
The invention changes the output parameters of the neural network from discrete point coordinates to B-spline control point parameters. By parameterizing the lane geometry using B-spline functions with control points, the network transforms discrete predictions into continuous curve representations through parameter transformation rather than geometric reconstruction
Solution Approach 2:
The invention replaces the mechanical/geometric approach of connecting discrete points with line segments with a mathematical parametric approach using B-spline functions. Instead of mechanically reconstructing lines from discrete points, the system uses parametric curve equations that inherently represent continuous geometry, substituting geometric reconstruction with parametric modeling
Data Source
AI summary
A method for training a model to detect lane markings. The method includes: providing images that are specific to a recording by at least one sensor of a vehicle, and in which the lane markings are mapped; providing ground truth data specific to a geometry of the lane markings in the provided images; training the model on the basis of the provided images and ground truth data for a three-dimensional modeling of the geometry of the lane markings based on a parameterization of a continuous curve.