3D Model Smoothing via Confidence-Based Artifact Reduction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for constructing three-dimensional models of geographic features often result in artifacts due to varying quality, leading to inaccurate representations that can be distracting when viewed from different vantage points.

Innovation Solution

A system determines confidence values based on the accuracy and eccentricity of model portions, using techniques such as stereo triangulation, residual errors, and orientation, to apply smoothing inversely proportional to confidence and eccentricity, thereby improving model quality and user experience.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If smoothing is applied to portions of the 3D model, then visual quality and accuracy are improved, but computational complexity and processing time increase

Engineering Contradiction:
Improvemodel accuracyVSAvoidprocessing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies smoothing selectively to specific portions of the 3D model based on confidence values and artifact detection, rather than uniformly processing the entire model. This local approach improves accuracy where needed while minimizing unnecessary computational overhead in already-quality regions.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The 3D model is divided into multiple portions or regions, each evaluated independently for quality metrics. This segmentation allows the system to identify and process only those portions requiring smoothing, reducing overall computational complexity while maintaining model accuracy.

Inventive Principle:
Principle #1Segmentation

2Reliability

If selective smoothing is applied based on confidence values, then artifact reduction is improved, but computational overhead increases

Engineering Contradiction:
Improveartifact reductionVSAvoidcomputational overhead
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system uses confidence values inherently generated during the photogrammetric processing to automatically identify regions requiring smoothing. This self-service approach leverages existing data structures without requiring separate complex analysis pipelines, reducing computational overhead while improving artifact reduction.

Inventive Principle:
Principle #25Self-service

3Measurement precision

If extensive smoothing is applied to improve model quality, then visual appeal is improved, but sharp features and details may be lost

Engineering Contradiction:
Improvevisual qualityVSAvoidfeature detail
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

Different smoothing intensities are applied to different portions of the model based on local feature importance and confidence metrics. This preserves sharp features in critical regions while applying stronger smoothing to low-confidence or less important areas, maintaining visual quality without information loss.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP3304500B1Smoothing 3D models of objects to mitigate artifacts
Publication Date: 2020.05.13 GOOGLE LLC
  • EP3304500B1 patent drawingFigure 1
  • EP3304500B1 patent drawingFigure 2
  • EP3304500B1 patent drawingFigure 3~4

AI summary

A system and method is provided for smoothing portions of a 3D model (250) of an object (208) based on the likelihood of a portion being an accurate representation of the surface of the object, and based on whether the surface of the object according to the model is relatively jagged (250) or relatively smooth (260).