Image Analysis Computing Device for Object Change Assessment

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current image analysis methods for assessing object changes are inaccurate due to variations in image acquisition conditions and material properties, leading to inconsistencies in comparing real-time images with older images.

Innovation Solution

An image analysis computing device and method that accounts for differences in image acquisition by using machine learning to classify and identify changes in objects made of various materials, employing techniques like keypoint detection and descriptor matching to improve accuracy and robustness against image position, alignment, and material variations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If current image comparison methods are used to assess object changes, then the process is simple and fast, but the accuracy is poor due to variations in image acquisition conditions and material properties

Engineering Contradiction:
Improveimage change assessment accuracyVSAvoidimage processing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the image comparison process into multiple stages: initial comparison to identify potential changes, material classification to determine material type, and secondary comparison using material-appropriate metrics. This segmentation allows the system to handle different materials with appropriate methods, improving accuracy without overwhelming complexity at any single stage

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes the comparison parameters based on material type. For metallic objects, it uses metrics sensitive to metallic surface changes; for non-metallic objects, it uses different metrics appropriate for those materials. This dynamic parameter adjustment resolves the contradiction by adapting the measurement approach to the specific material being examined

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If traditional image comparison is performed without accounting for material properties, then the processing is quick and simple, but the results are inaccurate due to different mechanical behaviors of materials

Engineering Contradiction:
Improvechange detection accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary material classification before executing the detailed change detection process. By identifying the material type early in the workflow, the system can select the appropriate comparison metrics and algorithms in advance, avoiding the need for time-consuming trial-and-error approaches and enabling accurate material-specific analysis

Inventive Principle:
Principle #10Preliminary action

3Measurement precision

If image position, alignment, and orientation variations are not corrected, then the comparison process is faster, but the assessment accuracy deteriorates

Engineering Contradiction:
Improveimage comparison accuracyVSAvoidprocessing efficiency
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent extracts and corrects position, alignment, and orientation variations as separate preprocessing steps before performing the actual change detection. By isolating these geometric corrections from the main comparison task, the system efficiently handles transformations without compromising the core analysis speed or accuracy

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP2911112B1Methods for assessing image change and devices thereof
Publication Date: 2019.04.17 WIPRO LTD
  • EP2911112B1 patent drawingFigure 1
  • EP2911112B1 patent drawingFigure 2
  • EP2911112B1 patent drawingFigure 3

AI summary

A method, non-transitory computer readable medium, and an image analysis computing device (14) that retrieves (302), based on a captured version of an object in a received image, training images which display related versions of the object and items of data related to the related versions of the object of the training images. Keypoints which are invariant to changes in scale and rotation in the captured version of the object in the received image and in the related versions of the object in the training images are determined (306). Changes to the object in the received image based on any of the determined keypoints in the related version of the object which do not match the determined keypoints in the captured version of the object are identified (316). The identified changes in the captured version of the object in the received image are provided (324).