Voxel Ground Plane Segmentation for 3D LiDAR Object Tracking

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Segmenting three-dimensional data from LIDAR systems to distinguish between ground planes, static, and dynamic objects in environments is challenging due to the complexity of the data and the need for efficient processing methods that can accurately identify and differentiate these elements for autonomous vehicle navigation.

Innovation Solution

The method involves representing LIDAR data in a voxel space, using techniques like region growing and ray casting to identify ground planes and objects, and distinguishing between static and dynamic objects by analyzing changes over time, allowing for efficient segmentation and processing that can be used for autonomous vehicle navigation and map generation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional image segmentation methods are applied to three-dimensional LIDAR data, then object identification can be achieved, but processing complexity and computational requirements increase significantly

Engineering Contradiction:
Improveobject identification accuracyVSAvoidprocessing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides the three-dimensional LIDAR data into discrete volumetric pixels (voxels) that can be independently processed. Each voxel represents a small volume element in the 3D space, allowing the system to process spatial data in a grid-based manner similar to 2D image processing, thereby reducing computational complexity while maintaining segmentation accuracy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extends traditional 2D image segmentation concepts into 3D space by introducing the voxel structure. This dimensional transformation allows the application of efficient 2D processing algorithms to each slice or layer of the 3D data, reducing the overall computational burden compared to processing the entire 3D point cloud directly.

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

2Measurement precision

If detailed segmentation of three-dimensional data is performed to distinguish ground planes and objects, then navigation accuracy improves, but processing time increases

Engineering Contradiction:
Improveground plane and object detection accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the 3D LIDAR data into voxels and further divides the processing into distinct stages: ground plane detection, object segmentation, and classification. This multi-stage segmentation approach allows the system to process different portions of the data with appropriate algorithms, reducing overall processing time while maintaining detection accuracy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs ground plane detection and removal as a preliminary step before object segmentation. By identifying and separating the ground plane first, the system reduces the complexity of subsequent object detection tasks, as objects can be detected in the remaining free space without the complexity of ground surface variations, thereby reducing total processing time.

Inventive Principle:
Principle #10Preliminary action

3Speed

If three-dimensional LIDAR data is processed in real-time for autonomous navigation, then safety and responsiveness improve, but computational load increases

Engineering Contradiction:
Improvereal-time processing speedVSAvoidcomputational load
Core Design Contradiction:
SpeedVSPower

Solution Approach 1:

The patent processes LIDAR data by dividing it into voxels and further segmenting into ground plane and object portions. This segmentation enables parallel processing of different voxel regions and reduces the computational load per processing unit, allowing real-time operation with reduced power requirements.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements a simplified voxel processing approach that processes only the necessary portions of the 3D data for navigation decisions. By using coarse-grained voxel representations and processing only relevant regions (e.g., areas with detected objects or changes), the system achieves real-time performance without processing the entire dataset at maximum detail, reducing computational load.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3639241B1Voxel based ground plane estimation and object segmentation
Publication Date: 2023.11.15 ZOOX INC
  • EP3639241B1 patent drawingFigure 1
  • EP3639241B1 patent drawingFigure 2
  • EP3639241B1 patent drawingFigure 3

AI summary

Systems, methods, and apparatuses described herein are directed to performing segmentation on voxels representing three-dimensional data to identify static and dynamic objects. LIDAR data may be captured by a perception system for an autonomous vehicle and represented in a voxel space. Operations may include determining a drivable surface by parsing individual voxels to determine an orientation of a surface normal of a planar approximation of the voxelized data relative to a reference direction. Clustering techniques can be used to grow a ground plane including a plurality of locally flat voxels. Ground plane data can be set aside from the voxel space, and the remaining voxels can be clustered to determine objects. Voxel data can be analyzed over time to determine dynamic objects. Segmentation information associated with ground voxels, static object, and dynamic objects can be provided to a tracker and/or planner in conjunction with operating the autonomous vehicle.