AR Hand Point-Cloud Rendering for Accurate Virtual Occlusion

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing AR gesture interaction methods fail to accurately render the occlusion relationship between hands and virtual objects, leading to unrealistic user interactions in augmented reality scenes.

Innovation Solution

Utilizing a combination of RGB images, depth images, and inertial measurement unit data to perform three-dimensional dense modeling, enabling accurate rendering of the occlusion relationship between hands and virtual objects through three-dimensional point cloud information and pose information.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If 2D SLAM model is used for gesture interaction, then device complexity is reduced, but occlusion relationship rendering between hands and virtual objects cannot be achieved

Engineering Contradiction:
Improvecomplexity of gesture interaction systemVSAvoidaccuracy of occlusion relationship rendering
Core Design Contradiction:
Device complexityVSManufacturing precision

Solution Approach 1:

The patent transitions from 2D SLAM model to 3D dense modeling by introducing depth dimension through RGB-D images and point cloud generation. This dimensional expansion enables accurate representation of hand geometry and occlusion relationships with virtual objects, resolving the limitation of 2D approaches而无法实现遮挡关系的问题

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

The patent combines multiple data sources (RGB images, depth images, IMU data) to create a composite 3D representation of hands through point cloud generation. This composite approach integrates visual and inertial information to achieve accurate occlusion rendering while maintaining system feasibility

Inventive Principle:
Principle #40Composite materials

2Manufacturing precision

If 3D dense modeling with RGB-D images is implemented, then occlusion relationship accuracy is improved, but use of energy increases

Engineering Contradiction:
Improveaccuracy of occlusion relationship renderingVSAvoidenergy consumption of gesture processing
Core Design Contradiction:
Manufacturing precisionVSUse of energy by moving object

Solution Approach 1:

The patent segments the hand modeling process into distinct stages: RGB-D image acquisition, point cloud generation, pose estimation, and occlusion rendering. This segmentation allows selective processing and optimization of energy-intensive operations, reducing overall energy consumption while maintaining accuracy

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements selective 3D dense modeling only for hand regions identified in the scene, rather than processing entire frames. This partial action approach reduces computational load and energy consumption while maintaining accurate occlusion rendering where needed

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3971686B1Ar scenario-based gesture interaction method and apparatus, storage medium, and communication terminal
Publication Date: 2025.11.19 GUANGDONG OPPO MOBILE TELECOMMUNICATIONS CORP LTD
  • EP3971686B1 patent drawingFigure 1
  • EP3971686B1 patent drawingFigure 2~3
  • EP3971686B1 patent drawingFigure 4~6

AI summary

Disclosed area an AR scenario-based gesture interaction method, an AR scenario-based gesture interaction apparatus, a computer-readable medium, and a communication terminal, relating to the field of augmented reality technology. Said method comprises: acquiring an RGB image and a depth image of a current frame, and corresponding IMU data (S11); processing the RGB image, the depth image and the IMU data so as to acquire attitude information and hand information of the current frame (S12); performing three-dimensional dense modeling according to the attitude information and the hand information of the current frame so as to acquire three-dimensional point cloud information of the hand in a reference coordinate system (S13); acquiring pose information of a virtual object in the reference coordinate system (S14); and on the basis of the three-dimensional point cloud information of the hand and the pose information of the virtual object in the reference coordinate system, and preset point cloud information of the virtual object, rendering the hand and the virtual object, so as to acquire a shielding relationship between the hand and the virtual object (S15).