Abstract :

In Modern Manufacturing Environments, Surface Defects In Metallic And Industrial Components Critically Affect Product Reliability, Safety, And Overall Production Efficiency. Traditional Inspection Methods, Predominantly Based On Manual Evaluation, Suffer From Inconsistencies, Subjectivity, And Increased Operational Costs. Earlier Automated Approaches Relying On Handcrafted Features Often Fail To Generalize Under Varying Lighting Conditions, Noise, And Complex Defect Patterns. To Address These Limitations, A Robust Deep Feature–driven Framework, Termed Defect Net, Was Developed By Integrating Transfer Learning With Multi-classifier Fusion. The Framework Employed The Exception Pre-trained Model As A Deep Feature Extractor To Capture High-level Discriminative Representations From Surface Images. Leveraging Depth Wise Separable Convolutions, The Model Efficiently Extracted Rich And Compact Features Suitable For Diverse Defect Categories. These Extracted Features Were Subsequently Utilized By Multiple Machine Learning Classifiers, Including Stochastic Gradient Descent (SGD), Passive Aggressive Classifier (PAC), Histogram-based Gradient Boosting (HGB), And Quadratic Discriminant Analysis (QDA), Enabling Comparative Analysis And Improved Classification Robustness. To Further Enhance Detection Performance, A Multi-scale Convolutional Neural Network (CNN) Model Was Incorporated As A Proposed Deep Learning Approach For End-to-end Classification. The Fusion Of Deep Features With Multiple Classifiers Significantly Improved Accuracy, Precision, And Generalization Capability Across Varying Defect Types. Additionally, The System Was Deployed Within A User-friendly Graphical Interface, Facilitating Real-time Defect Prediction Through Simple Image Uploads. Experimental Evaluations Demonstrated That The Proposed Framework Achieved Superior Performance Compared To Conventional Methods, Ensuring Reliable And Scalable Defect Detection. This Approach Contributes To Intelligent Quality Assurance Systems By Reducing Human Dependency And Enabling Efficient Real-time Industrial Inspection.

Published:

10-4-2026

Issue:

Vol. 26 No. 4 (2026)

Page Nos:

2210-2220

Section:

Articles

License:

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

How to Cite