Lidar Point Cloud Annotation via Registration

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The generation of large volumes of high-quality annotated datasets for training machine-learning algorithms, particularly for autonomous driving applications, is time-consuming and costly due to the need for extensive manual annotation of lidar data.

Innovation Solution

A method that aligns and annotates lidar point clouds across multiple datasets using a point-cloud registration algorithm to transform them into a globally consistent coordinate space, reducing the number of annotation steps and increasing efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If manual annotation of lidar data is performed extensively to generate high-quality annotated datasets, then the quality and size of the dataset improve, but the time and cost required to obtain the data increase significantly

Engineering Contradiction:
Improveannotation qualityVSAvoidannotation time
Core Design Contradiction:
Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent applies preliminary action by performing point cloud registration and alignment before the annotation process. Multiple lidar point clouds are pre-processed to establish spatial relationships and temporal consistency, creating a unified coordinate framework. This preliminary structuring of data enables automated or semi-automated annotation processes, significantly reducing the manual time required while maintaining high annotation quality.

Inventive Principle:
Principle #10Preliminary action

2Manufacturing precision

If manual annotation of lidar data is performed extensively to generate high-quality annotated datasets, then the quality and size of the dataset improve, but the cost required to obtain the data increases significantly

Engineering Contradiction:
Improveannotation qualityVSAvoidannotation cost
Core Design Contradiction:
Manufacturing precisionVSEase of manufacture

Solution Approach 1:

The patent replaces the mechanical manual annotation process with an automated computational system. By implementing point cloud registration algorithms and automated annotation tools that operate on pre-processed lidar data, the system eliminates the need for extensive manual human labor. This substitution dramatically reduces annotation costs while preserving data quality through consistent, error-free automated processing.

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

3Measurement precision

If point clouds from multiple lidar output datasets are aligned into a globally consistent coordinate space, then the accuracy of object tracking and annotation improves, but the computational complexity and processing time increase

Engineering Contradiction:
Improvespatial alignment accuracyVSAvoidprocessing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies segmentation by dividing the point cloud processing into distinct modular stages: first associating point clouds across datasets, then performing registration to align them, and finally annotating the aligned results. This segmentation allows each computational task to be optimized independently and enables parallel processing of multiple point clouds, reducing overall computational complexity while achieving high spatial alignment accuracy.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP4485394A1Annotation of object states in lidar output datasets
Publication Date: 2025.01.01 ZENSEACT AB
  • EP4485394A1 patent drawingFigure 1
  • EP4485394A1 patent drawingFigure 2
  • EP4485394A1 patent drawingFigure 3

AI summary

A method for generating an annotated dataset for training and/or verifying machine-learning algorithms and related aspects are disclosed. The method comprises associating point clouds across a plurality of lidar output datasets in order to link each point cloud across the plurality of lidar output datasets that represent the same object, where the plurality of lidar output datasets comprise point-cloud representations of one or more objects over a time period. The method further comprises, for each object of the one or more objects, aligning the associated point clouds representing the same object across the plurality of lidar output datasets using a point-cloud registration algorithm configured to spatially transform the point cloud of one or more lidar output datasets in order to align the associated point clouds representing the same object across the plurality of lidar output datasets into a globally consistent coordinate space. The method further comprises, for each object of the one or more objects, annotating the aligned point cloud, and converting the annotation to an original coordinate space of the plurality of lidar output datasets so to obtain an annotation for each lidar output dataset of the plurality of lidar output datasets.