Multi-Channel Defect Detection Model with Merging Module
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional manual visual inspection and semi-automatic optical instrument-assisted quality inspection in industrial manufacturing often lead to misdetected defects, and there is an inconvenience in storing and utilizing inspection results effectively.
Innovation Solution
A product defect detection method utilizing a multi-channel image acquisition and a defect detection model with convolutional branches and a merging module to perform feature extraction and merge characteristic information across channels, enhancing defect detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual visual inspection is used, then operation simplicity is maintained, but detection precision deteriorates due to misdetected defects
Solution Approach 1:
The patent replaces manual visual inspection with an automated optical detection system that captures multi-channel images and processes them through convolutional neural network branches. This substitution eliminates human error in defect detection while maintaining operational simplicity through automated processing pipelines.
Solution Approach 2:
The detection system is segmented into multiple independent convolutional branches, each specialized for processing specific image channels (e.g., RGB channels, depth channel, normal map channel). This segmentation allows each branch to optimize feature extraction for its specific channel type, improving overall detection precision without creating a monolithic complex system.
2Measurement precision
If multi-channel image processing is implemented, then detection precision improves through enhanced feature extraction, but device complexity increases due to multiple convolutional branches and merging modules
Solution Approach 1:
The complex detection task is divided into multiple specialized convolutional branches, each handling specific image channels. This segmentation transforms a potentially overwhelming monolithic model into manageable, specialized components that can be independently optimized and trained.
Solution Approach 2:
Multiple convolutional branches processing different image channels are merged through a merging module that integrates their feature extractions. This merging combines the strengths of each branch while maintaining a structured, organized architecture that avoids the chaos of uncoordinated complexity.
3Loss of information
If conventional inspection methods are used, then system simplicity is maintained, but information utilization efficiency deteriorates due to inconvenience in storing and utilizing inspection results
Solution Approach 1:
The system creates digital copies of inspection results in the form of multi-channel images and extracted feature data. These digital copies can be stored, transmitted, and re-analyzed without degrading the information, enabling efficient archival and future utilization of inspection data.
Data Source
AI summary
A product defect detection method and apparatus, an electronic device, and a storage medium are provided. A method includes: acquiring a multi-channel image of a target product; inputting the multi-channel image to a defect detection model, wherein the defect detection model includes a plurality of convolutional branches, a merging module and a convolutional headbranch; performing feature extraction on each channel in the multi-channel image by using the plurality of convolutional branches, to obtain a plurality of first characteristic information; merging the plurality of first characteristic information by using the merging module, to obtain second characteristic information; performing feature extraction on the second characteristic information by using the convolutional headbranch, to obtain third characteristic information to be output by the defect detection model; and determining defect information of the target product based on the third characteristic information.