Camera-LiDAR Alignment for Road Plane and Obstacle Navigation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Autonomous vehicles face challenges in navigating complex road environments due to misalignment between LIDAR and camera systems, which affects the accurate correlation of laser reflections with visual data, leading to difficulties in determining road planes and obstacle detection.

Innovation Solution

A system that estimates a road plane using optical flow from camera images and correlates this with LIDAR reflections to attribute laser information to objects in the environment, enabling accurate navigation on curved or obstructed roads.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If the system correlates LIDAR reflections with camera images to attribute laser information to objects, then navigation accuracy on curved or obstructed roads is improved, but the complexity of processing and aligning multi-sensor data increases

Engineering Contradiction:
Improvenavigation reliabilityVSAvoiddata processing complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system segments the point cloud data into distinct regions corresponding to different object types (vehicles, pedestrians, road surfaces, obstacles) based on spatial clustering and feature analysis. This segmentation allows independent processing of different object categories, reducing overall computational complexity while maintaining navigation reliability.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system transforms 3D LIDAR point cloud data into 2D image space using the calibrated transformation matrices, allowing correlation with 2D camera images. This dimensional transformation enables efficient attribute assignment by mapping volumetric laser reflections onto planar image features where object detection algorithms can operate more efficiently.

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

2Adaptability or versatility

If the system uses optical flow from camera images to estimate road plane, then the ability to navigate on varied road surfaces is improved, but the accuracy is reduced when roads are not flat or are curved

Engineering Contradiction:
Improveroad condition adaptabilityVSAvoidroad plane estimation accuracy
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The system merges road plane estimation results from both camera-based optical flow analysis and LIDAR-based 3D surface reconstruction. By combining these complementary approaches, the system leverages the wide-field view of cameras for broad road surface coverage while using LIDAR's precise depth measurements to correct for curvature and non-flat conditions, achieving both versatility and accuracy.

Inventive Principle:
Principle #5Merging (Combining)

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

Enhances the ability of autonomous vehicles to navigate by accurately determining road planes and obstacle locations, improving safety and efficiency in various road conditions.

Implementation Method 1

receive an output of a LIDAR onboard the host vehicle, wherein the output of the LIDAR is representative of a plurality of laser reflections

Methodology Applied
Scientific EffectLIDAR: LIDAR

Implementation Method 2

the output of the LIDAR is representative of a plurality of laser reflections from at least an area associated with a first portion of the road

Methodology Applied
Scientific EffectLaser reflection: Reflection

Implementation Method 3

estimate a road plane representative of the first portion of the road based on the captured images, using optical flow across the captured images

Methodology Applied
Scientific EffectOptical flow:

Data Source

PatentEP3545337B1Vehicle navigation based on aligned image and lidar information
Publication Date: 2024.07.03 MOBILEYE VISION TECH LTD
  • EP3545337B1 patent drawingFigure 1
  • EP3545337B1 patent drawingFigure 2A
  • EP3545337B1 patent drawingFigure 2B

AI summary

Systems and methods are provided for navigating an autonomous vehicle. In one implementation, a navigational system for a host vehicle may include at least one processor programmed to: receive a stream of images captured by a camera onboard the host vehicle, wherein the captured images are representative of an environment surrounding the host vehicle; and receive an output of a LIDAR onboard the host vehicle, wherein the output of the LIDAR is representative of a plurality of laser reflections from at least a portion of the environment surrounding the host vehicle. The at least one processor may also be configured to determine at least one indicator of relative alignment between the output of the LIDAR and at least one image captured by the camera; attribute LIDAR reflection information to one or more objects identified in the at least one image based on the at least one indicator of the relative alignment between the output of the LIDAR and the at least one image captured by the camera; and use the attributed LIDAR reflection information and the one or more objects identified in the at least one image to determine at least one navigational characteristic associated with the host vehicle.