Landmark Identification from Ground-Level Point Clouds

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current geographic information systems lack an efficient method to identify prominent landmarks from geographic imagery data captured at ground level, which are visible from multiple perspectives and at greater distances, limiting their effectiveness in providing accurate travel directions and rendering in emphasized styles.

Innovation Solution

A computer-implemented method that generates a point cloud from geographic imagery data using structure-from-motion techniques, filters data points based on distance from camera perspectives, and matches them with known landmarks, prioritizing them based on visibility scores calculated from viewsheds.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional landmark identification methods are used, then the system can identify some landmarks, but it cannot efficiently identify prominent landmarks that are visible from multiple perspectives and at greater distances

Engineering Contradiction:
Improvelandmark identification accuracyVSAvoidlandmark identification efficiency
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent segments the landmark identification process into distinct stages: generating a point cloud from imagery data, filtering points by distance from camera perspectives, calculating viewshed metrics for each point, and ranking points based on visibility scores. This segmentation allows each stage to be optimized independently, improving both accuracy and efficiency.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary actions by pre-calculating viewsheds and visibility metrics for all candidate points before final landmark selection. This allows the system to efficiently identify prominent landmarks by having visibility data already prepared, rather than computing it during the selection process.

Inventive Principle:
Principle #10Preliminary action

2Reliability

If the system processes all data points in the point cloud, then comprehensive landmark identification is achieved, but computational complexity and processing time increase significantly

Engineering Contradiction:
Improvelandmark identification completenessVSAvoidsystem computational complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent extracts and processes only the most relevant data points by filtering the point cloud based on distance from camera perspectives. This extraction of key points (those visible from multiple perspectives and at greater distances) reduces the dataset size while maintaining identification reliability.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies different processing criteria to different regions of the point cloud based on their spatial relationship to camera perspectives. Points that are farther from camera perspectives and visible from multiple angles are given higher priority, creating a localized quality assessment rather than uniform processing.

Inventive Principle:
Principle #3Local quality

3Measurement precision

If the system identifies landmarks visible from multiple perspectives and distances, then travel direction accuracy is improved, but the complexity of analyzing multiple camera perspectives increases

Engineering Contradiction:
Improvetravel direction accuracyVSAvoidmulti-perspective analysis complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent introduces viewshed analysis as an intermediary mechanism that simplifies multi-perspective analysis. Instead of directly analyzing all camera perspectives, the system uses viewshed calculations to aggregate visibility information, creating a simplified metric that represents multi-perspective visibility without requiring complex direct analysis of each perspective.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP3044726B1Landmark identification from point cloud generated from geographic imagery data
Publication Date: 2024.05.22 GOOGLE LLC
  • EP3044726B1 patent drawingFigure 1
  • EP3044726B1 patent drawingFigure 2
  • EP3044726B1 patent drawingFigure 3~4

AI summary

Systems and methods for identifying prominent landmarks in a geographic area are provided. More particularly, geographic imagery data, such as panoramic images captured by a camera from a perspective at or near ground level, can be analyzed to identify landmarks depicted in the geographic imagery data. For instance, structure-from-motion techniques can be used to generate a point cloud of tracked features that are located a threshold distance away from one or more camera perspectives associated with the geographic imagery data. Landmarks that are visible from multiple camera perspectives and that are visible from greater distances relative to one or more camera perspectives can be identified as prominent landmarks in a geographic area.