AI-Based FOD Detection With Explainable Image Captions

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current FOD detection techniques require manual inspection, lack visual information, and need a standard reference for comparison, failing to provide detailed natural language descriptions for easier interpretation.

Innovation Solution

An FOD detection system utilizing generative AI to generate text captions for captured images, combined with machine learning algorithms to identify FOD and an explainable AI model for detailed location understanding, providing natural language translations for easier interpretation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual inspection is used for FOD detection, then detailed visual information can be obtained, but the detection process is time-consuming and labor-intensive

Engineering Contradiction:

Improvedetailed visual informationVSAvoiddetection process time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces manual mechanical inspection with an automated system combining imaging devices, generative AI models, and text classifiers. The imaging devices capture images of the inspection area, the generative AI model creates detailed natural language descriptions, and the text classifier identifies FOD objects, thereby eliminating manual labor while maintaining detailed visual information capture

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system enables self-service detection by using the imaging devices and AI models to automatically perform the inspection task without human intervention. The generative AI model independently generates descriptions and the text classifier automatically identifies FOD, making the detection process autonomous and eliminating the need for manual inspection time

Inventive Principle:

Principle #25Self-service

2Productivity

If existing automated detection techniques are used, then detection speed is improved, but they lack natural language descriptions for easier interpretation

Engineering Contradiction:

Improvedetection speedVSAvoidnatural language descriptions

Core Design Contradiction:

ProductivityVSLoss of information

Solution Approach 1:

The patent introduces a generative AI model as an intermediary between the image capture and FOD detection processes. This intermediary translates visual data into natural language descriptions, providing both the speed of automated detection and the interpretability of human-like language explanations about detected FOD objects

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system changes the output parameter from simple detection signals to detailed natural language descriptions. The generative AI model transforms image data into descriptive text that explains what FOD objects are detected, their locations, and characteristics, thereby adding interpretability information without sacrificing detection speed

Inventive Principle:

Principle #35Parameter changes

3Measurement precision

If standard reference comparison is used for FOD detection, then detection accuracy is improved, but the system requires pre-existing reference data that may not be available

Engineering Contradiction:

Improvedetection accuracyVSAvoidavailability of reference data

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent replaces the reference-comparison mechanism with a generative AI description mechanism. Instead of comparing images against stored references, the system generates natural language descriptions of detected objects and uses text classification to identify FOD, thereby achieving accurate detection without requiring pre-existing reference data for every possible FOD type

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS20250391184A1Automated Foreign Object Debris Detection System Using Generative AI

Publication Date: 2025.12.25 THE BOEING CO

AI summary

Detecting foreign object debris (FOD) is provided. The method comprises receiving images captured by a number of imaging devices in a defined inspection area. A generative artificial intelligence (AI) model generates a natural language text caption describing the images. A text classifier AI model classifies FOD in the images based on the natural language text caption. An explainable AI model identifies words and phrases within the natural language text caption according to which the text classifier AI model made the classification. A report is displayed in a user interface, wherein for each identified FOD the report includes a location where the FOD was identified in the inspection area, the captured image with the natural language text caption, and highlighting of the key words and phrases identified by the explainable AI model. A generative AI translation model can then translate the report into a specified second language.