LIDAR Point Cloud Annotation for Fast 3D Box and Road Surface Estimation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Manual annotation of 3-D LIDAR point clouds for vehicular driver assist and autonomous driving systems is time-consuming, expensive, and prone to errors, and the detection of unusual stationary objects poses data scalability issues, requiring efficient and accurate annotation methods.

Innovation Solution

The system provides fast and accurate annotation cluster pre-proposals on a minimally-supervised or unsupervised basis by estimating 3-D bounding boxes and drivable surfaces using LIDAR point clouds, employing algorithms for feature-based detection of similar objects in already-annotated frames to reduce ambiguity and training manual annotators.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual annotation of 3-D LIDAR point clouds is performed, then accurate annotated data is obtained, but time consumption and expense increase significantly

Engineering Contradiction:
Improveannotation accuracyVSAvoidannotation time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary automated annotation to generate pre-proposals for bounding boxes and drivable surfaces before manual annotation. This preliminary action provides a head start for annotators, reducing the time needed to create accurate annotations while maintaining high precision through manual refinement of the automated results

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

An automated algorithm acts as an intermediary between raw LIDAR data and manual annotation. The algorithm generates pre-proposals that serve as intermediate results, which annotators then refine. This intermediary step reduces the direct time burden on annotators while preserving annotation accuracy through human oversight

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If manual annotation of unusual stationary objects is performed, then complete object detection is achieved, but expertise requirements and expense increase

Engineering Contradiction:
Improveobject detection completenessVSAvoidannotation expertise requirement
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system enables self-service annotation by automatically detecting unusual stationary objects and generating pre-proposals for them. This reduces the need for expert annotators to manually identify rare objects, as the automated system handles the initial detection and proposal generation, requiring only verification and refinement by annotators

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent replaces the mechanical process of expert manual annotation with an automated computational system. The algorithm substitutes human expertise in detecting unusual objects by using feature-based detection and comparison with already-annotated frames, reducing dependency on specialized annotator knowledge

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Productivity

If feature-based detection of similar objects is implemented, then annotation scalability is improved, but algorithm complexity increases

Engineering Contradiction:
Improveannotation scalabilityVSAvoidalgorithm complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The system uses copying by detecting similar objects in already-annotated frames and reusing their annotation patterns. When a new object is detected, the algorithm copies the annotation approach from previously annotated similar objects, enabling scalable annotation without requiring complex manual intervention for each new object type

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The system performs preliminary feature extraction and object similarity comparison before final annotation. By pre-identifying similar objects and their annotation patterns, the system prepares the groundwork for scalable annotation, reducing the complexity burden during the actual annotation process

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3644276B1Methods and systems for the fast estimation of three-dimensional bounding boxes and drivable surfaces using lidar point clouds
Publication Date: 2022.05.04 VOLVO CAR CORP
  • EP3644276B1 patent drawingFigure 1
  • EP3644276B1 patent drawingFigure 2
  • EP3644276B1 patent drawingFigure 3

AI summary

The present invention relates to methods and systems for generating annotated data for training vehicular driver assist (DA) and autonomous driving (AD) active safety (AS) functionalities and the like. More specifically, the present invention relates to methods and systems for the fast estimation of three-dimensional (3-D) bounding boxes and drivable surfaces using LIDAR point clouds and the like. These methods and systems provide fast and accurate annotation cluster pre-proposals on a minimally-supervised or unsupervised basis, segment drivable surfaces/ground planes in a bird's-eye-view (BEV) construct, and provide fast and accurate annotation cluster pre-proposal labels based on the feature-based detection of similar objects in already-annotated frames. The methods and systems minimize the expertise, time, and expense associated with the manual annotation of LIDAR point clouds and the like in the generation of annotated data for training machine learning (ML) algorithms and the like.