3D Lane Generation Using Neural Depth and Spatial Offsets

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing deep learning-based lane recognition technologies struggle to accurately identify and generate three-dimensional lane information in diverse road environments, lacking effective methods for integrating topographical features and depth information.

Innovation Solution

A neural network-based approach that generates lane probability maps, feature information, depth distributions, and spatial information to produce accurate three-dimensional lane data by applying multiple neural networks for lane feature extraction, depth estimation, and spatial transformation using calibration parameters.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If deep learning-based lane recognition is applied to diverse road environments, then lane recognition potential increases, but accuracy in identifying topographical features deteriorates

Engineering Contradiction:
Improvelane recognition potentialVSAvoidaccuracy in identifying topographical features
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent transforms 2D lane detection into 3D lane information by introducing depth estimation through neural networks. The system generates three-dimensional lane data including height and distance information, moving from planar image space to volumetric space to capture topographical features accurately

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

The patent divides the lane detection task into multiple independent neural network components: one network for lane probability mapping, another for depth estimation, and a third for 3D lane generation. This segmentation allows each component to specialize in specific features while maintaining overall system adaptability

Inventive Principle:
Principle #1Segmentation

2Measurement precision

If multiple neural networks are used for feature extraction and depth estimation, then lane recognition accuracy improves, but system complexity increases

Engineering Contradiction:
Improvelane recognition accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent combines multiple neural network outputs (lane probability maps and depth estimates) into a unified 3D lane representation. By merging these separate feature extractions into a single volumetric output, the system achieves high accuracy while managing complexity through integrated processing

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP4177850B1Method and apparatus with lane generation
Publication Date: 2025.07.09 SAMSUNG ELECTRONICS CO LTD
  • EP4177850B1 patent drawingFigure 1
  • EP4177850B1 patent drawingFigure 2
  • EP4177850B1 patent drawingFigure 3

AI summary

A method of generating lane information using a neural network includes generating a lane probability map based on an input image, generating lane feature information and depth feature information by applying the lane probability map to a second neural network, generating depth distribution information by applying the depth feature information to a third neural network, generating spatial information based on the lane feature information and the depth distribution information, generating offset information including a displacement between a position of a lane and a reference line by applying the spatial information to a fourth neural network, and generating three-dimensional (3D) lane information using the offset information.