Single-Scan LiDAR Object Identification Using Pre-Trained ML Classifier

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing LiDAR systems mounted on vehicles face challenges in achieving high precision object identification with a reduced number of scan lines, leading to increased calculation costs and potential recognition rate degradation due to height differences in sensor mounting.

Innovation Solution

A vehicular object identification system using a single scan line with a distance sensor and a classifier trained by machine learning, utilizing training data from a LiDAR with multiple scan lines to improve data acquisition efficiency and reduce calculation load, allowing object recognition independent of sensor height.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If the number of scan lines is reduced to lower calculation costs, then device complexity and processing requirements are reduced, but object identification precision deteriorates

Engineering Contradiction:
Improveprocessing device complexityVSAvoidobject identification precision
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent applies preliminary action by pre-training a machine learning classifier using high-resolution multi-scan line LiDAR data before deployment. The classifier learns object identification patterns from comprehensive training data, enabling it to achieve high identification accuracy when processing only single-scan line data during actual vehicle operation, thus resolving the contradiction between reduced scan lines and maintained precision

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent uses copying by creating a trained machine learning model that captures the essential object identification knowledge from multi-scan line LiDAR data. This model copy can then operate efficiently with reduced scan line input, transferring the learning benefits from comprehensive data to simplified real-time processing

Inventive Principle:
Principle #26Copying

2Ease of manufacture

If a low-cost processing device is used, then device cost is reduced, but calculation capability and object identification precision worsen

Engineering Contradiction:
Improveprocessing device costVSAvoidobject identification precision
Core Design Contradiction:
Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent employs cheap short-living objects by using a low-cost processing device that performs efficient single-scan line classification. The system accepts reduced computational capabilities in exchange for using pre-trained models that require minimal real-time calculation, making the processing device more affordable while maintaining acceptable identification performance

Inventive Principle:
Principle #27Cheap short-living objects (Disposable)

Solution Approach 2:

The patent applies mechanics substitution by replacing complex real-time calculation mechanisms with a pre-trained machine learning classifier. Instead of relying on high computational power during operation, the system uses knowledge embedded in the trained model, allowing low-cost hardware to achieve reliable object identification

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

3Productivity

If single scan line data is used for object identification, then data processing speed and device simplicity are improved, but recognition accuracy deteriorates due to insufficient information

Engineering Contradiction:
Improveobject identification speedVSAvoidobject identification accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies parameter changes by transforming the approach from using multiple scan lines (spatial parameter) to using a single scan line enhanced by machine learning parameters. The classifier learns to extract sufficient identification information from the single-scan-line spatial constraint by adjusting and optimizing learning parameters during training, thereby maintaining accuracy while improving processing speed

Inventive Principle:
Principle #35Parameter changes

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

Enables efficient and cost-effective object identification using a single scan line, reducing processing requirements and enhancing recognition accuracy despite varying sensor heights, while maintaining low-cost hardware configurations.

Implementation Method 1

a distance sensor configured to scan a single beam in a horizontal direction so as to measure distances to points on a surface of an object

Methodology Applied
Scientific EffectLight reflection: Reflection

Implementation Method 2

LiDAR (Light Detection and Ranging) including a plurality of scan lines in a vertical direction

Methodology Applied
Scientific EffectLight detection and ranging: LIDAR

Data Source

PatentEP3889646B1On-vehicle object identification system, automobile, vehicle lamp, classifier learning method, and arithmetic operation device
Publication Date: 2025.06.25 KOITO MFG CO LTD
  • EP3889646B1 patent drawingFigure 1
  • EP3889646B1 patent drawingFigure 2
  • EP3889646B1 patent drawingFigure 3

AI summary

A vehicular object identification system 10 includes a distance sensor 20 and a processing device 40. The distance sensor 20 scans a single beam in the horizontal direction so as to measure the distances to points P on the surface of an object OBJ. The processing device 40 includes a classifier 42 that is capable of identifying the kind of the object OBJ based on point cloud data PCD that corresponds to the single scan line acquired by the distance sensor 20. The classifier 42 is implemented based on a learned model generated by machine learning. The machine learning is executed using multiple items of point cloud data that correspond to multiple scan lines acquired by measuring a predetermined object by means of a LiDAR that supports the multiple scan lines in the vertical direction.