3D Model Reconstruction for Deformable Objects via Depth Image Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing 3D model scanning techniques struggle to accurately reconstruct deformable objects due to structural misalignment caused by changes during the scanning process, and the use of multiple cameras increases costs, making it unfeasible for ordinary users.

Innovation Solution

A method and apparatus that divide depth images of a deformable object into groups based on linking information, building local models for each substructure and merging them to create an accurate 3D model, using a single camera and reducing the need for multiple cameras.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If a single camera is used to shoot the object at different time points, then the cost is reduced, but the reconstructed model is incomplete due to structural misalignment caused by appearance changes during shooting

Engineering Contradiction:
ImprovecostVSAvoidmodel accuracy
Core Design Contradiction:
Ease of manufactureVSManufacturing precision

Solution Approach 1:

The patent divides the depth images into multiple groups based on linking information that records location information of substructures. Each group corresponds to a specific substructure, allowing independent processing and reconstruction of different parts of the object. This segmentation enables the system to track and reconstruct deformable objects accurately by maintaining correspondence between substructures across different time points, thereby resolving the contradiction between using a single camera (low cost) and achieving accurate reconstruction (high precision).

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent handles dynamic changes in object appearance during the shooting period by using linking information to track substructures across different time points. The system dynamically updates the reconstruction by maintaining the correspondence relationships between substructures, allowing accurate reconstruction of deformable objects even when their appearance changes during the scanning process. This dynamic approach enables single-camera systems to achieve model accuracy comparable to multi-camera systems.

Inventive Principle:
Principle #15Dynamics

2Manufacturing precision

If multiple cameras are used to simultaneously shoot the object, then the model accuracy is improved, but the required cost is higher than using a single camera

Engineering Contradiction:
Improvemodel accuracyVSAvoidcost
Core Design Contradiction:
Manufacturing precisionVSEase of manufacture

Solution Approach 1:

The patent segments the depth images into groups corresponding to different substructures, allowing a single camera to capture and process different parts of the object at different time points. By using linking information to track substructure locations, the system achieves the spatial coverage and accuracy normally requiring multiple simultaneous cameras, but with the cost advantage of using only one camera. The segmentation enables temporal substitution for spatial multiplication.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent transitions from a spatial approach (multiple cameras simultaneously capturing the object) to a temporal approach (single camera capturing the object at different time points). By adding the time dimension and using linking information to track substructures across time, the system achieves the reconstruction accuracy of multi-camera systems while using a single camera, effectively substituting temporal data for spatial data.

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

3Device complexity

If a single camera is used to shoot the object at different time points, then the device complexity is reduced, but the structural misalignment causes incomplete reconstruction

Engineering Contradiction:
Improvecamera configurationVSAvoidstructure alignment
Core Design Contradiction:
Device complexityVSManufacturing precision

Solution Approach 1:

The patent segments the object into substructures and groups depth images accordingly, using linking information to track the location of each substructure across different time points. This segmentation enables the system to maintain structural alignment by processing and reconstructing each substructure independently while maintaining their spatial relationships, thereby achieving accurate reconstruction with simple single-camera hardware. The segmentation compensates for the lack of simultaneous multi-view data by organizing temporal data into spatially coherent groups.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent uses linking information as feedback to track and correct structural alignment issues. The linking information records the location information of substructures across different time points, providing feedback that enables the system to detect and correct misalignment. This feedback mechanism allows the single-camera system to maintain structural coherence by continuously tracking and adjusting the reconstruction based on the recorded substructure positions, thereby achieving high precision without complex hardware.

Inventive Principle:
Principle #23Feedback

Data Source

PatentUS9262862B2Method and apparatus for reconstructing three dimensional model
Publication Date: 2016.02.16 IND TECH RES INST
  • US9262862B2 patent drawing
  • US9262862B2 patent drawing
  • US9262862B2 patent drawing

AI summary

A method and an apparatus for reconstructing a three dimensional model of an object are provided. The method includes the following steps. A plurality of first depth images of an object are obtained. According to a linking information of the object, the first depth images are divided into a plurality of depth image groups. The linking information records location information corresponding to a plurality of substructures of the object. Each depth image group includes a plurality of second depth images, and the substructures correspond to the second depth images. According to the second depth image and the location information corresponding to each substructure, a local module of each substructure is built. According to the linking information, the local models corresponding to the substructures are merged, and the three-dimensional model of the object is built.