Defect Detection System Adaptive Training Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Precision machine vision systems face challenges in accurately detecting defects in workpieces due to variations in workpiece types and surfaces, as well as changing inspection conditions, necessitating an improvement in defect detection accuracy and efficiency.
Innovation Solution
A workpiece inspection system that includes a light source, lens, camera, and processors, utilizing training images to train a defect detection model, which determines the accuracy performance and adjusts the training based on the performance metrics to optimize the detection of defects, allowing for continuous improvement in defect detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional machine vision systems are used for workpiece inspection, then the system structure is simple, but the defect detection accuracy is insufficient due to variations in workpiece types and surfaces
Solution Approach 1:
The system performs preliminary training by collecting defect images and training data before actual inspection operations. This preliminary action prepares the neural network model in advance, enabling it to adapt to different workpiece types and surfaces, thereby improving detection accuracy without increasing operational complexity
Solution Approach 2:
The system dynamically adjusts training parameters including the number of training images, defect image collection requirements, and performance thresholds based on detected accuracy levels. This adaptive parameter adjustment optimizes the balance between detection accuracy and system complexity by only collecting additional images when performance falls below thresholds
2Measurement precision
If more training images are collected to improve detection accuracy, then the accuracy performance improves, but the time and resources required for training increase
Solution Approach 1:
The system implements a feedback mechanism that continuously monitors defect detection accuracy and compares it against predefined performance thresholds. Based on this feedback, the system automatically determines whether additional training images are needed, preventing unnecessary image collection and training time while ensuring accuracy requirements are met
Solution Approach 2:
The system collects training images in staged portions rather than all at once. It initially collects a baseline set of images, trains the model, evaluates performance, and only collects additional images if accuracy thresholds are not met. This partial action approach minimizes unnecessary image collection and training time while achieving required accuracy
3Adaptability or versatility
If the system adapts to different workpiece types and inspection conditions, then the versatility improves, but the complexity of maintaining consistent detection accuracy increases
Solution Approach 1:
The system performs preliminary training with defect images specific to each workpiece type and inspection condition before deployment. This preliminary adaptation ensures the model learns the characteristics of different workpieces in advance, maintaining consistent detection accuracy across variations without requiring complex real-time adjustments
Solution Approach 2:
The system dynamically adjusts training parameters such as the number of required defect images and performance thresholds based on the specific workpiece type and inspection conditions being evaluated. This adaptive parameter adjustment maintains detection consistency across different scenarios by optimizing training requirements for each specific case
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
The system enhances defect detection accuracy by dynamically adjusting the training data based on performance metrics, improving the model's ability to classify defects and non-defects, thereby increasing the efficiency and reliability of the inspection process.
Implementation Method 1
a light source, a lens, a camera... The lens inputs image light arising from a surface of a workpiece which is illuminated by the light source
Implementation Method 2
The lens inputs image light arising from a surface of a workpiece which is illuminated by the light source, and transmits the image light along an imaging optical path
Implementation Method 3
The camera receives imaging light transmitted along the imaging optical path and provides an image of the workpiece
Data Source
AI summary
A workpiece inspection and defect detection system includes a light source, a lens that inputs image light arising from a surface of a workpiece, and a camera that receives imaging light transmitted along an imaging optical path. The system utilizes images of workpieces acquired with the camera as training images to train a defect detection portion to detect defect images that include workpieces with defects, and determines a performance of the defect detection portion as trained with the training images. Based on the performance of the defect detection portion, an indication is provided as to whether additional defect images should be provided for training. After training is complete, the camera is utilized to acquire new images of workpieces which are analyzed to determine defect images that include workpieces with defects, and for which additional operations may be performed (e.g., metrology operations for measuring dimensions of the defects, etc.)