Autoencoder Defect Detection Noise Removal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing defect detection methods using autoencoders are prone to inaccuracies due to noise in test samples, requiring direct comparison and potentially leading to incorrect defect determination.
Innovation Solution
A device and method employing a trained autoencoder and pixel convolutional neural network that generates a test encoding feature by removing noise from the test sample, allowing for defect detection without direct comparison, using trained weightings from defect-free training samples to output a probability-based result.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If direct comparison between test sample and constructed image is used for defect determination, then the defect detection process is simple, but noise in the test sample causes inaccurate defect determination
Solution Approach 1:
The patent extracts only the essential features of the test sample by encoding it into a test encoding feature using the autoencoder, separating the essential information from noise. This extracted encoding is then compared with the encoded defect-free sample, effectively removing noise influence while maintaining detection accuracy.
Solution Approach 2:
The patent introduces an intermediary representation (encoding feature) as a mediator between the original test sample and the defect determination process. This intermediary encoding captures the essential characteristics while filtering out noise, enabling accurate defect detection without directly comparing noisy original images.
2Measurement precision
If noise is removed from test sample before defect detection, then defect determination accuracy improves, but the processing complexity increases
Solution Approach 1:
The patent performs preliminary encoding of both defect-free training samples and test samples into encoding features before defect determination. This preliminary action transforms the data into a noise-resistant representation space, simplifying the subsequent defect detection process while improving accuracy.
Solution Approach 2:
The patent replaces traditional mechanical noise filtering methods with a learned encoding transformation through autoencoder. This substitution uses neural network-based feature extraction to automatically identify and preserve essential features while ignoring noise, achieving noise removal without complex manual processing.
Data Source
AI summary
In a method for defecting surface defects, a trained weighting generated when defect-free training samples are used to train an autoencoder and pixel convolutional neural network is obtained. A test encoding feature is obtained by inputting the trained weighting into the autoencoder and pixel convolutional neural network and a weighted autoencoder of the weighted autoencoder and pixel convolutional neural network encoding a test sample. The test encoding feature is input into a weighted pixel convolution neural network of the weighted autoencoder and pixel convolutional neural network to output a result of test. The test result is either no defect in the test sample or at least one defect in the test sample. Inaccurate determinations as to defects are thereby avoided. An electronic device and a non-transitory storage medium are also disclosed.