Deep Learning Machine Vision for Rail Vehicle Inspection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current machine vision inspection systems face challenges in automating the inspection of complex environments, such as railyards, due to difficulties in processing images of moving objects in varying lighting conditions and the need for human intervention, which is costly and hazardous.

Innovation Solution

A deep learning-based machine vision system that uses a combination of graphics processing units, vision processing units, and tensor processing units to analyze image data, identify regions of interest, and determine inspection outcomes, with the option for human feedback and retraining to improve performance over time.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If human inspectors manually inspect rail vehicles, then inspection accuracy can be maintained through human judgment, but labor costs increase and safety hazards arise

Engineering Contradiction:

Improveinspection accuracyVSAvoidlabor cost

Core Design Contradiction:

ReliabilityVSEase of manufacture

Solution Approach 1:

The patent replaces the mechanical human inspection system with an automated machine vision system using cameras, processors, and algorithms to detect and classify rail vehicle components, eliminating manual labor costs while maintaining inspection accuracy through computational analysis

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The inspection system performs self-validation through automated image analysis and classification algorithms that independently evaluate rail vehicle components without requiring human intervention, enabling the system to serve itself in making inspection decisions

Inventive Principle:

Principle #25Self-service

2Productivity

If human inspectors manually inspect moving rail vehicles, then inspection can be performed in real-time, but safety hazards increase due to exposure to moving equipment

Engineering Contradiction:

Improvereal-time inspection capabilityVSAvoidsafety hazard

Core Design Contradiction:

ProductivityVSObject-affected harmful factors

Solution Approach 1:

The patent replaces human inspectors with automated imaging devices and processing systems that can safely inspect moving rail vehicles without exposing personnel to hazardous environments, maintaining real-time inspection capability while eliminating safety risks

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent introduces an intermediary automated vision system that acts as a mediator between the inspection requirement and the moving rail vehicle, allowing real-time inspection to occur at a safe distance without direct human exposure to moving equipment

Inventive Principle:

Principle #24Intermediary (Mediator)

3Ease of manufacture

If machine vision systems are used to automate inspection, then labor costs decrease and safety improves, but the system struggles to process complex images in varying lighting conditions

Engineering Contradiction:

Improvelabor cost reductionVSAvoidimage processing complexity

Core Design Contradiction:

Ease of manufactureVSDifficulty of detecting and measuring

Solution Approach 1:

The patent segments the complex inspection task into distinct processing stages: image capture, preprocessing, feature extraction, and classification. This segmentation allows each stage to handle specific aspects of image processing independently, reducing overall complexity while maintaining accuracy in varying lighting conditions

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies preliminary image preprocessing operations such as normalization, contrast enhancement, and noise reduction before main analysis. This preliminary action prepares images for subsequent processing steps, making them more robust to lighting variations and reducing the difficulty of detecting features in complex conditions

Inventive Principle:

Principle #10Preliminary action

4Measurement precision

If deep learning models are implemented for automated inspection, then inspection accuracy improves, but system complexity and computational requirements increase

Engineering Contradiction:

Improveinspection accuracyVSAvoidsystem complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent implements a universal deep learning model that can classify multiple types of rail vehicle components (bogies, trucks, couplers, etc.) using a single trained system. This multi-functionality reduces the need for separate specialized systems for each component type, managing complexity while maintaining high inspection accuracy across diverse components

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS20220172335A1Machine Vision Based Inspection

Publication Date: 2022.06.02 INTERNATIONAL ELECTRONIC MACHINES CORP

AI summary

A solution for inspecting one or more objects of an apparatus. An inspection component obtains an inspection outcome for an object of the apparatus based on image data of the apparatus. The inspection component can use a deep learning engine to analyze the image data and identify object image data corresponding to a region of interest for the object. A set of reference equipment images can be compared to the identified object image data to determine the inspection outcome for the object. The inspection component can further receive data regarding the apparatus, which can be used to determine a general location of the object on the apparatus and therefore a general location of the region of interest for the object in the image data. The inspection component can be configured to provide image data in order to obtain feedback from a human and/or further train the deep learning engine.