Deep Learning Sheet Inspection for Low-Contrast Defect Sorting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sheet part inspection systems struggle to accurately detect low contrast and low frequency defects in manufactured films, leading to potential misclassification of quality, which can result in discarded high-quality parts or shipment of low-quality parts.
Innovation Solution
A system utilizing convolutional neural networks (CNNs) trained to identify specific defects in sheet parts, combined with a classifier to determine quality categories based on defect presence and rules, enabling accurate sorting into satisfactory, defective, or rework categories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional image processing techniques are used for inspection, then the system is simpler and faster, but it cannot accurately detect low contrast and low frequency defects
Solution Approach 1:
The patent replaces traditional mechanical image processing techniques with deep learning neural networks to detect defects. The neural networks process image data to identify low contrast and low frequency defects that traditional methods miss, achieving higher measurement precision through computational intelligence rather than conventional optical or mechanical inspection mechanisms
Solution Approach 2:
The patent changes the detection parameters by using multiple neural networks with different training focuses. Each neural network is trained to detect specific types of defects with different characteristics, allowing the system to adapt to various defect patterns and improve overall detection accuracy across different defect types
2Adaptability or versatility
If multiple types of defects are detected using a single inspection system, then the system becomes more versatile, but the complexity of the inspection device increases
Solution Approach 1:
The patent segments the defect detection function into multiple separate neural networks, where each network is specialized for detecting specific types of defects. This segmentation allows each network to be optimized for its particular defect type while the overall system maintains versatility through the combination of multiple specialized networks
Solution Approach 2:
The patent creates a universal inspection system by combining multiple neural networks that can detect different defect types. The system processes the same image data through multiple networks simultaneously, enabling a single inspection device to handle diverse defect characteristics and provide comprehensive quality assessment
Data Source
AI summary
An inspection system includes an inspection device having at least one image capture device. The image capture device captures image data of a sheet part passing through the inspection device. A processing unit of the inspection device provides the image data representative of the sheet part to a plurality of neural networks, where each of the neural networks is trained to identify a corresponding defect in the sheet part and output data indicative of the presence of the corresponding defect. The processing unit determines a quality category of the sheet part based on the data indicative of the presence of the corresponding defect output by each corresponding neural network. The processing unit can further output the quality category of the sheet part to a sorter that can sort the sheet part based on the quality category.