3D Image Frame Confidence Scoring for Element Recognition

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image processing techniques struggle with identifying elements in complex 3-dimensional objects, particularly when specific views of these elements are required, and there is a need for improved methods to enhance the accuracy and efficiency of element recognition in 3D imaging.

Innovation Solution

A machine learning algorithm, specifically a convolutional neural network (CNN), is trained to recognize elements in specified planes of 3D objects, providing confidence levels and saliency maps to assist operators in identifying and capturing suitable image frames, with features like bounding boxes and color changes to highlight elements, and a system for training using batches of image frames with balanced background and foreground elements.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If machine learning algorithms are used to identify elements in 3D objects, then recognition accuracy is improved, but device complexity increases

Engineering Contradiction:
Improverecognition accuracyVSAvoiddevice complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent introduces confidence levels as an intermediary metric between the machine learning algorithm and the operator. Instead of directly presenting complex algorithmic outputs, the system translates recognition results into intuitive confidence level indicators that guide operator decision-making, thereby managing the complexity-accuracy tradeoff

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If confidence levels are allocated to each image frame, then element identification accuracy is improved, but processing time increases

Engineering Contradiction:
Improveelement identification accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system calculates confidence levels for all image frames but strategically selects only those frames exceeding threshold values for operator review and storage. This partial action approach processes comprehensive data while minimizing actual operator workload and effective processing time

Inventive Principle:
Principle #16Partial or excessive action

3Reliability

If multiple image frames are stored with confidence levels, then data quality is improved, but storage requirements increase

Engineering Contradiction:
Improvedata qualityVSAvoidstorage requirements
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

The patent applies local quality by storing image frames with differentiated quality attributes based on their confidence levels. High-confidence frames are stored with full quality for definitive identification, while lower-confidence frames may be stored with reduced quality or metadata only, optimizing storage efficiency while maintaining data reliability where most needed

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP3520021B1Image processing
Publication Date: 2026.03.25 KINGS COLLEGE LONDON
  • EP3520021B1 patent drawingFigure 1
  • EP3520021B1 patent drawingFigure 2
  • EP3520021B1 patent drawingFigure 3

AI summary

Imaging methods, imaging apparatus and computer program products are disclosed. An imaging method comprises: receiving image data of a 3-dimensional object; and allocating a confidence level to at least a portion of an image frame of the image data using a machine-learning algorithm, the confidence level indicating a likelihood of that image frame having a specified element imaged on a specified plane through the 3- dimensional object. In this way, particular elements when imaged in a desired way can be identified from image data of the 3-dimensional object.