LiDAR Semantic Label Confidence Scoring for Reliable Point Selection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Accurate semantic label predictions for LiDAR point clouds remain an open problem, often dependent on the detail of the underlying ontology, and existing systems lack effective methods to evaluate the reliability of label assignments for downstream tasks.

Innovation Solution

A system utilizing both supervised and unsupervised machine learning algorithms to generate confidence scores for LiDAR data point labels, combining these scores to select reliable data points and aggregate them into a 3D representation for improved object detection and path planning.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If semantic labels are assigned to LiDAR data points using existing systems, then labeling can be performed, but the reliability of label assignments cannot be effectively evaluated

Engineering Contradiction:
Improvelabel assignment reliabilityVSAvoidconfidence information
Core Design Contradiction:
ReliabilityVSLoss of information

Solution Approach 1:

The system implements feedback by using the confidence estimation model to evaluate label assignments and provide confidence scores back to the labeling process. This feedback loop enables continuous improvement of label quality by identifying and correcting low-confidence assignments, thereby resolving the contradiction between achieving reliable labels and maintaining information completeness.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The confidence estimation model acts as an intermediary between the semantic labeling process and downstream tasks. It mediates by evaluating label assignments and providing confidence scores without requiring manual verification, thus enabling reliable label evaluation while preserving all original data information for further processing.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If all LiDAR data points are used for downstream tasks, then data quantity is maximized, but accuracy decreases due to inclusion of unreliable labels

Engineering Contradiction:
Improvedownstream task accuracyVSAvoiddata point quantity
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The system applies local quality by differentiating between individual data points based on their confidence scores. Instead of treating all data points uniformly, it selectively weights or filters points based on local confidence assessments, thereby improving overall accuracy by ensuring high-quality labels are prioritized while maintaining sufficient data quantity through confidence-threshold-based selection.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The system changes the parameter of data point selection by introducing confidence scores as an additional filtering criterion. By adjusting confidence thresholds dynamically, the system can optimize the balance between data quantity and accuracy for different downstream tasks, resolving the contradiction between maximizing data usage and ensuring label reliability.

Inventive Principle:
Principle #35Parameter changes

3Reliability

If confidence estimation is implemented using only supervised learning, then label consistency is maintained, but the system cannot evaluate reliability of labels from unlabeled or weakly labeled data

Engineering Contradiction:
Improvelabel reliability evaluationVSAvoiddata source flexibility
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The confidence estimation system achieves universality by designing a multi-functional model that can process and evaluate labels from multiple data sources including fully labeled, partially labeled, and unlabeled LiDAR data. This enables the system to adapt to different data availability scenarios while maintaining reliable confidence assessments, resolving the contradiction between label reliability and data source flexibility.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system implements self-service by enabling unlabeled or weakly labeled data to contribute to confidence estimation without requiring manual annotation. The model automatically evaluates the reliability of self-generated or externally sourced labels, allowing the system to utilize diverse data sources while maintaining quality control through automated confidence assessment.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP4148599B1Systems and methods for providing and using confidence estimations for semantic labeling
Publication Date: 2025.10.29 ARGO AI LLC
  • EP4148599B1 patent drawingFigure 1
  • EP4148599B1 patent drawingFigure 2
  • EP4148599B1 patent drawingFigure 3

AI summary

Systems and methods for processing and using sensor data. The methods comprise: obtaining semantic labels assigned to data points; performing a supervised machine learning algorithm and an unsupervised machine learning algorithm to respectively generate a first confidence score and a second confidence score for each semantic label of said semantic labels, the first and second confidence scores each representing a degree of confidence that the semantic label is correctly assigned to a respective one of the data points; generating a final confidence score for each said semantic label based on the first and second confidence scores; selecting subsets of the data points based on the final confidence scores; and aggregating the data points of the subsets to produce an aggregate set of data points.