Range Image Dimensioning via Selective Viewpoint Extraction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional dimensioning systems in material handling environments face challenges in efficiently measuring large objects due to the need for multiple image captures, leading to high memory usage, processing time, and error-prone image registration, which reduces throughput and accuracy.

Innovation Solution

The system captures and processes a reduced number of range images by selecting viewpoints based on pre-determined criteria, such as geometric features, to compute object dimensions without 3D image stitching, using a pattern projector and range camera to identify 3D points and estimate correspondence scores for accurate dimension calculation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If multiple image captures are used to measure large objects, then measurement coverage is improved, but processing time increases

Engineering Contradiction:
Improvemeasurement coverageVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent extracts only the necessary viewpoint information from multiple images by identifying geometric features and computing correspondences between selective viewpoints. Instead of processing all captured images, the system extracts essential 3D point correspondence data, significantly reducing processing time while maintaining measurement coverage for large objects.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent segments the image processing task by dividing it into discrete geometric feature identification and correspondence computation steps. By breaking down the complex task of measuring large objects into manageable segments (feature detection, viewpoint selection, correspondence estimation), the system achieves efficient processing without sacrificing measurement accuracy.

Inventive Principle:
Principle #1Segmentation

2Measurement precision

If multiple image captures are used to measure large objects, then measurement coverage is improved, but memory usage increases

Engineering Contradiction:
Improvemeasurement coverageVSAvoidmemory usage
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The system extracts only essential 3D point correspondence data from captured images rather than storing and processing entire image datasets. By taking out only the necessary geometric feature correspondences needed for dimension calculation, memory usage is significantly reduced while maintaining the ability to measure large objects accurately.

Inventive Principle:
Principle #2Taking out (Extraction)

3Measurement precision

If multiple image captures are used to measure large objects, then measurement coverage is improved, but registration accuracy decreases

Engineering Contradiction:
Improvemeasurement coverageVSAvoidregistration accuracy
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent replaces traditional mechanical image registration systems with a geometric feature-based correspondence estimation approach. By substituting complex registration mechanics with direct geometric feature matching and 3D point correspondence computation, the system achieves accurate dimension measurement without the errors inherent in traditional image registration methods.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

4Loss of information

If conventional image processing is used, then comprehensive object data is captured, but throughput decreases

Engineering Contradiction:
Improveobject data completenessVSAvoidthroughput
Core Design Contradiction:
Loss of informationVSProductivity

Solution Approach 1:

The system extracts only the essential 3D point correspondence information needed for dimension calculation from captured images. By taking out only the critical geometric data required for measuring length, width, and height, the system maintains complete object dimension data while dramatically improving processing throughput for material handling environments.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies partial action by computing correspondences only for selective viewpoints and geometric features rather than processing all image data comprehensively. This partial processing approach captures sufficient object dimension information to calculate length, width, and height accurately while enabling high-speed throughput necessary for productive material handling operations.

Inventive Principle:
Principle #16Partial or excessive action

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

This approach significantly reduces processing time and improves accuracy by capturing fewer images while maintaining high throughput, enabling efficient dimension calculation of large objects in material handling environments.

Implementation Method 1

capturing multiple image frames of an object from a plurality of locations... retrieving a plurality of three-dimensional (3D) point clouds from the plurality of captured image frames

Methodology Applied
Scientific EffectOptical triangulation: Parallax

Data Source

PatentEP3588436B1Methods, systems, and apparatuses for computing dimensions of an object using range images
Publication Date: 2023.11.29 HAND HELD PRODS INC
  • EP3588436B1 patent drawingFigure 1
  • EP3588436B1 patent drawingFigure 2
  • EP3588436B1 patent drawingFigure 3

AI summary

Various embodiments described herein relate to techniques for computing dimensions of an object using multiple range images. In this aspect, the multiple range images are captured from selective locations and satisfy a pre-defined criterion. In accordance with various embodiments, at least a pair of 3D points are identified from the multiple range images, which correspond to at least one geometric feature on a surface of the object. In this regard, a correspondence score is estimated for the identified at least one pair of 3D points. The correspondence score is then utilized for registering the at least one pair of 3D points. Based in part on the registration of the 3D points and 3D point clouds retrieved from the captured images, the dimensions of the object are computed.