Advancing workpiece dimension measurement: Integrating AI-based edge detection with machine vision and coordinate...

Area of Science:

• Industrial Metrology
• Computer Vision
• Artificial Intelligence

Background:

• Accurate workpiece dimension measurement is crucial in industrial manufacturing.
• Interference regions on surfaces pose challenges for traditional edge detection and roundness assessment.
• Existing machine vision systems often struggle with complex surface features and multi-hole measurements.

Purpose of the Study:

• To investigate the application of AI-based detection methods within industrial image analysis for enhanced measurement accuracy.
• To develop and validate a Convolutional Neural Network (CNN) for identifying interference regions and predicting circularity.
• To compare the performance of the AI approach against Coordinate Measuring Machines (CMMs) and conventional vision systems.

Main Methods:

• Design and fabrication of two physical models with varying hole configurations using CNC milling.
• Application of a transfer learning approach with a CNN for image analysis and defect detection.
• Image preprocessing techniques including multi-layer convolution and pooling for feature enhancement.
• Comparative diameter analysis and statistical validation using ANOVA.

Main Results:

• The CNN achieved 100% classification accuracy in identifying interference regions.
• All measurement methods maintained a deviation of ≤0.05 mm from actual values.
• The CNN matched CMM precision (r=1.000) and outperformed conventional vision systems in multi-hole measurements.
• The method demonstrated cross-material scalability after retraining.

Conclusions:

• Integrating deep learning techniques, specifically CNNs, significantly enhances precision in industrial inspection and dimension measurement.
• The developed AI-driven methodology offers a standardized approach for accurate workpiece analysis.
• This research validates the synergy of machine vision, deep learning, and CMMs for advancing industrial measurement capabilities.