Image Inspection Controller with Iterative Classification

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current image inspection systems often rely on single classification approaches, lack flexibility, and do not allow user input during the classification process, leading to suboptimal output and difficulty in identifying process failures.

Innovation Solution

The system integrates semantic segmentation and object classification processes, allowing user input and iteration between these processes to generate a more detailed output image, utilizing traditional machine learning techniques that require limited training data and computing resources.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If multiple classification processes are applied in a pre-defined order, then the classification output is generated, but the system lacks flexibility and user input capability

Engineering Contradiction:

ImproveflexibilityVSAvoidsystem complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The system dynamically adjusts the classification process based on user input and intermediate results. Users can select which classification process to apply next (pixel-level or object-level) based on the current output quality, making the system adaptive rather than following a fixed predetermined sequence. This resolves the contradiction by enabling flexibility without requiring a completely complex reconfigurable architecture.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system incorporates feedback loops where users can review intermediate classification outputs and provide input to guide subsequent processing steps. This feedback mechanism allows users to determine when additional classification is needed and what type, providing flexibility while maintaining manageable system complexity through structured interaction protocols.

Inventive Principle:

Principle #23Feedback

2Measurement precision

If only one classification process is used, then the system is simple to operate, but the output image lacks detail

Engineering Contradiction:

Improveoutput detailVSAvoidoperational simplicity

Core Design Contradiction:

Measurement precisionVSEase of operation

Solution Approach 1:

The classification process is segmented into distinct stages (pixel-level classification and object-level classification), where each stage contributes different types of detail to the output. Users can selectively apply additional classification stages based on their needs, achieving high output detail while maintaining operational simplicity through on-demand processing rather than mandatory multi-stage execution.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system allows users to apply classification processes partially or iteratively rather than requiring complete multi-stage processing for all cases. Users can stop after achieving sufficient detail, avoiding unnecessary complexity while ensuring high detail is available when needed. This resolves the contradiction by making detailed classification available without mandating complex operational procedures.

Inventive Principle:

Principle #16Partial or excessive action

3Adaptability or versatility

If iterative classification with user input is implemented, then user customization is enabled, but the process time increases

Engineering Contradiction:

Improveuser customizationVSAvoidprocess time

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

The iterative classification process uses periodic user feedback points where users can review outputs and decide whether to continue processing. This structured periodic interaction enables user customization while controlling process time by providing clear decision points rather than continuous or unbounded iteration. Users can approve outputs at any stage, preventing unnecessary time consumption.

Inventive Principle:

Principle #19Periodic action

4Use of energy by moving object

If traditional machine learning techniques are used, then computing resources are reduced, but the training data requirements are also reduced

Engineering Contradiction:

Improvecomputing resourcesVSAvoidtraining data

Core Design Contradiction:

Use of energy by moving objectVSQuantity of substance

Solution Approach 1:

The system uses traditional machine learning models that are computationally efficient and can be quickly trained and discarded if needed, rather than relying on complex deep learning models requiring extensive computing resources and large training datasets. This approach treats the classification models as lightweight, replaceable components that consume minimal resources, resolving the contradiction by making both computing resources and training data requirements low through deliberate model selection.

Inventive Principle:

Principle #27Cheap short-living objects (Disposable)

Data Source

PatentUS12205362B2Systems and methods for image classification

Publication Date: 2025.01.21 MOLECULAR DEVICES LLC

AI summary

In one aspect, a method for inspecting features of an image using an image inspection controller that includes a processor communicatively coupled to a memory is described. The method includes receiving, at the processor, an input image, performing, on the input image, one of a semantic segmentation process and an object classification process to generate an output image, and prompting a user to select between approving the displayed output image, and at least one of i) performing an additional semantic segmentation process on the displayed output image, and ii) performing an additional object classification process on the displayed output image.