LiDAR Intensity Segmentation for Occlusion-Free Depth Datasets

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for creating high-density depth datasets using LiDAR are limited by low resolution and require additional sensors like stereo cameras, leading to inaccurate depth data and erroneous removal of point clouds, especially in areas without texture.

Innovation Solution

An information processing apparatus that accumulates LiDAR point clouds over multiple frames, performs image segmentation based on reflection intensity, and removes occlusion point clouds using the segmentation results to create a high-density, high-accuracy depth dataset.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If LiDAR point clouds for multiple frames are accumulated to achieve higher density, then point cloud density is improved, but occlusion point clouds cannot be accurately distinguished and removed

Engineering Contradiction:
Improvepoint cloud densityVSAvoidocclusion removal accuracy
Core Design Contradiction:
Quantity of substanceVSReliability

Solution Approach 1:

The patent applies image segmentation to the intensity image obtained from LiDAR to divide the scene into different regions. By segmenting the intensity image based on reflection intensity values, the system can identify occlusion areas and distinguish them from visible areas, enabling accurate removal of occlusion point clouds while preserving the high-density accumulated point clouds.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent uses the intensity image as an intermediary between the accumulated point clouds and the occlusion removal process. The intensity image, which contains reflection intensity information, serves as a mediator to identify occlusion regions, allowing the system to accurately remove occlusion point clouds without requiring additional sensors like stereo cameras.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If stereo camera is added to obtain stereo depth data for occlusion removal, then occlusion detection capability is improved, but system complexity and size increase

Engineering Contradiction:
Improveocclusion detection capabilityVSAvoidsystem complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent makes the LiDAR system multi-functional by using it not only for obtaining depth information through point clouds but also for obtaining intensity images that contain reflection intensity information. This intensity image serves multiple purposes: it helps identify occlusion regions and guides the removal of occlusion point clouds, eliminating the need for separate stereo cameras while maintaining occlusion detection capability.

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

Solution Approach 2:

The patent enables the LiDAR system to serve itself by utilizing its own intensity image output for occlusion detection and removal. Instead of requiring external stereo cameras, the system uses its inherent intensity information to identify and remove occlusion point clouds, making the system self-sufficient and reducing overall system complexity.

Inventive Principle:
Principle #25Self-service

3Reliability

If stereo depth data obtained by SGM is used for occlusion removal, then occlusion area identification is improved, but depth accuracy deteriorates especially in areas without texture

Engineering Contradiction:
Improveocclusion area identificationVSAvoiddepth accuracy
Core Design Contradiction:
ReliabilityVSMeasurement precision

Solution Approach 1:

The patent replaces the complex and inaccurate stereo depth data obtained by SGM with a simpler and more accurate approach: using the intensity image from LiDAR directly for occlusion identification. The intensity image provides sufficient information to identify occlusion regions without requiring the computationally intensive and accuracy-limited SGM algorithm, thereby maintaining depth accuracy even in areas without texture.

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

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

The method enhances the accuracy and density of depth datasets by accurately identifying and removing occlusion point clouds, enabling more precise depth estimation and 3D modeling, even in areas without texture, while allowing for a compact system design.

Implementation Method 1

an intensity image generation unit that generates an intensity image in which a pixel value represents a peak value or an accumulated value of reflection intensity from an environment obtained by the LiDAR

Methodology Applied
Scientific EffectLight reflection: Reflection

Data Source

PatentEP4682826A1Information processing device, information processing method, and program
Publication Date: 2026.01.21 SONY GROUP CORP
  • EP4682826A1 patent drawingFigure 1
  • EP4682826A1 patent drawingFigure 2
  • EP4682826A1 patent drawingFigure 3

AI summary

The present disclosure relates to an information processing apparatus, an information processing method, and a program that enable a high-density and high-accuracy depth dataset to be created more suitably. An accumulation unit accumulates point clouds for multiple frames acquired by a LiDAR, a segmentation unit performs image segmentation on an intensity image based on reflection intensity from an environment acquired by the LiDAR, and an occlusion removal unit removes an occlusion point cloud corresponding to an occlusion area from the accumulated point clouds by using an execution result of the image segmentation. The technology according to the present disclosure can be applied to an in-vehicle sensor fusion system.