3D Model Generation Using Combined Mask Images

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The visual volume intersection method for generating three-dimensional models of objects is prone to defects when structures within the image capturing scene obstruct parts of the target object, leading to inaccurate silhouettes and reduced geometric accuracy of the three-dimensional model.

Innovation Solution

A system that acquires area information for both the object and potential obstructing structures, generating three-dimensional shape data by integrating this information to create a high-quality three-dimensional model, even when parts of the object are hidden by structures.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If the visual volume intersection method is used to generate three-dimensional models, then the three-dimensional shape can be estimated from multi-viewpoint images, but defects occur in the model when structures obstruct parts of the target object

Engineering Contradiction:
Improvegeometric accuracy of three-dimensional modelVSAvoidcompleteness of silhouette information
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The system performs preliminary identification of structures that may obstruct the target object before generating the three-dimensional model. By detecting and marking these obstructing structures in advance, the system can exclude their corresponding visual volumes from the intersection calculation, preventing defects in the final model while maintaining geometric accuracy from unobstructed viewpoints.

Inventive Principle:
Principle #10Preliminary action

2Reliability

If mask images with lost silhouette portions are discarded to avoid defects, then model quality is maintained, but the geometric accuracy is reduced due to loss of available information

Engineering Contradiction:
Improvequality of three-dimensional modelVSAvoidgeometric accuracy
Core Design Contradiction:
ReliabilityVSMeasurement precision

Solution Approach 1:

The system extracts and removes the visual volumes corresponding to obstructing structures from the set of visual volumes used in the intersection method. By separating the obstructing structure information from the target object information, the system can utilize silhouette data from unobstructed regions while excluding defective regions, thereby maintaining both model quality and geometric accuracy without discarding useful information.

Inventive Principle:
Principle #2Taking out (Extraction)

3Productivity

If all visual volumes are used in the intersection calculation, then more information is utilized, but defects are introduced when obstructing structures are included

Engineering Contradiction:
Improveutilization of available dataVSAvoidquality of three-dimensional model
Core Design Contradiction:
ProductivityVSManufacturing precision

Solution Approach 1:

The system performs preliminary identification and classification of visual volumes into obstructing structure volumes and target object volumes before the intersection calculation. This preliminary action enables selective utilization of visual volumes, incorporating data from unobstructed regions while excluding defective regions, thereby maximizing productive use of available data without introducing defects into the final model.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentUS11798233B2Generation device, generation method and storage medium for three-dimensional model that remove a portion of the three-dimensional model
Publication Date: 2023.10.24 CANON KK
  • US11798233B2 patent drawing
  • US11798233B2 patent drawing
  • US11798233B2 patent drawing

AI summary

A generation device of a three-dimensional model including: an acquisition unit configured to acquire a first mask image indicating a structure area, which is an object still within each image captured from a plurality of viewpoints, and a second mask image indicating a foreground area, which is a moving object within each image captured from the plurality of viewpoints; a combination unit configured to generate a third mask image that integrates the structure area and the foreground area within the image captured from the plurality of viewpoints by combining the first mask image and the second mask image both acquired; and a generation unit configured to generate a three-dimensional model including the structure and the foreground by a visual volume intersection method using the third mask image.