Defect Detection via Luminance Profile Frequency Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing defect inspection techniques struggle with accurately detecting scratches and other abnormalities on surfaces, particularly when the light source environment changes or shadows are present, and often require complex calibration or prior learning with databases.
Innovation Solution
An information processing apparatus that captures images of a target object illuminated from multiple directions, generates luminance profiles to classify defects without prior learning, and distinguishes between shadows and highlights, allowing for the detection of shape, hue, and gloss defects without the need for complex calibration or databases.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple light sources are used to detect various defects, then the types of detectable defects increase, but the device complexity and calibration requirements increase
Solution Approach 1:
The patent segments the defect detection task by analyzing luminance profiles at different spatial frequencies. Low-frequency components detect gloss defects, medium-frequency components detect shape defects, and high-frequency components detect hue defects. This segmentation allows a single illumination system to detect multiple defect types without increasing device complexity
Solution Approach 2:
The patent transforms the defect detection problem from spatial domain to frequency domain by applying Fourier transform to luminance profiles. This dimensional transformation enables simultaneous detection of multiple defect types through frequency component analysis, avoiding the need for multiple physical light sources and complex calibration
2Adaptability or versatility
If multiple light sources are used to detect various defects, then the types of detectable defects increase, but the calibration time increases
Solution Approach 1:
The system performs self-calibration by automatically analyzing the luminance profile characteristics of the target object under the illumination unit. The Fourier transform-based frequency component analysis automatically adapts to the specific object being inspected, eliminating the need for manual calibration with master data or databases
Solution Approach 2:
The patent changes the analysis parameter from spatial luminance values to frequency components of luminance profiles. This parameter transformation enables the system to detect different defect types by analyzing different frequency bands, eliminating the need for time-consuming calibration with multiple light sources
3Ease of operation
If conventional defect detection is used, then the detection process is simple, but shadows are erroneously detected as defects
Solution Approach 1:
The patent analyzes only specific frequency components of the luminance profile rather than the entire signal. By focusing on particular frequency bands characteristic of actual defects, the system distinguishes true defects from shadows while maintaining operational simplicity
Solution Approach 2:
The system uses Fourier transform to provide frequency-domain feedback about the luminance profile characteristics. This feedback mechanism enables the system to identify patterns characteristic of real defects versus shadows, improving detection reliability without significantly complicating the detection process
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables accurate detection of multiple abnormal states as defects on unknown objects and in varying light conditions, avoiding false positives from shadows, and simplifies the setup process by not requiring precise calibration of light sources.
Implementation Method 1
an image group obtained by imaging a target object irradiated with light from a plurality of illumination units placed at a plurality of different positions
Data Source
AI summary
An information processing apparatus for, to classify and detect a plurality of types of defects, detecting a defect in a target object without using data obtained by prior learning includes an acquisition unit configured to acquire an image group obtained by capturing a target object irradiated with light from a plurality of illumination units placed at a plurality of different positions, a generation unit configured to, based on luminance information of images included in the image group, generate a first luminance profile indicating a change in a luminance value of a first region, which is a region corresponding between the images, and a second luminance profile indicating a change in a luminance value of a second region, which is a region corresponding between the images, and a detection unit configured to detect a defect in the target object based on the first luminance profile and the second luminance profile.