3D Structure Classification via Surface Normal Variance

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current 3D graphics systems in navigation and mapping applications lack the ability to accurately classify and display 3D objects in virtual environments, leading to confusion between different structures such as roads, buildings, and statues, which is essential for effective navigation and context understanding.

Innovation Solution

An in-vehicle information system that classifies 3D structures by analyzing polygon data, including surface normals and dihedral angles, to identify and display structures as roads, regular buildings, irregular buildings, or statues, using a processor with integrated CPU and GPU capabilities to generate and display classified models with appropriate textures.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If 3D objects are displayed without classification information, then the display process is simple and fast, but the system cannot distinguish between different structures such as roads, buildings, and statues

Engineering Contradiction:
Improveclassification accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the classification process into distinct analytical steps: extracting geometric features from polygon data, calculating surface normals, determining dihedral angles, and comparing against classification thresholds. This segmentation enables systematic classification without overwhelming system complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary classification analysis by pre-defining geometric feature thresholds for different structure types (roads, buildings, statues). This preliminary preparation enables fast classification during runtime without complex real-time computation.

Inventive Principle:
Principle #10Preliminary action

2Loss of information

If classification analysis is performed on all 3D objects, then accurate structure identification is achieved, but processing time and computational resources increase

Engineering Contradiction:
Improveclassification information completenessVSAvoidprocessing time
Core Design Contradiction:
Loss of informationVSLoss of time

Solution Approach 1:

The patent extracts only the essential geometric features needed for classification (surface normals and dihedral angles) from the complete polygon data, rather than analyzing all possible properties. This extraction achieves accurate classification while minimizing processing time.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies partial analysis by focusing on specific geometric characteristics (orthogonality of surfaces) that are sufficient for classification purposes, rather than performing exhaustive analysis of all object properties. This partial action achieves the needed classification information efficiently.

Inventive Principle:
Principle #16Partial or excessive action

3Loss of information

If detailed classification of all structures is provided, then complete navigation information is available, but the display becomes cluttered and harder to interpret

Engineering Contradiction:
Improvenavigation information completenessVSAvoiddisplay interpretability
Core Design Contradiction:
Loss of informationVSEase of operation

Solution Approach 1:

The patent applies different display treatments to different classified structures based on their type. Roads receive one level of visual representation, buildings receive another, and statues receive yet another. This local differentiation provides complete navigation information while maintaining clear visual hierarchy and interpretability.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP2973428B1System and method for classification of three-dimensional models in a virtual environment
Publication Date: 2020.08.05 ROBERT BOSCH GMBH
  • EP2973428B1 patent drawingFigure 1
  • EP2973428B1 patent drawingFigure 2
  • EP2973428B1 patent drawingFigure 2

AI summary

A method for classification of three-dimensional structures in a virtual environment includes identifying a plurality of polygons in a structure located in a virtual environment, identifying a plurality of surface normals, each surface normal corresponding to one polygon in the plurality of polygons that are in the structure, identifying a variance of a distribution of the plurality of surface normals, generating a first classification for the structure in response to the variance being less than a predetermined threshold, and generating a graphical display of the structure with at least one visual aspect of the structure being modified with reference to the first classification.