Endüstriyel Döküm Parçalarda Yüksek Doğruluklu Yüzey Hata Sınıflandırması için Çoklu Özellik Çıkarımı ve Seçimi Temelli Bir Çerçeve

Öz

Bu çalışma, endüstriyel döküm parçalarının yüzey kusurlarının otomatik olarak tespit edilmesine yönelik görüntü işleme ve makine öğrenmesi tabanlı bir yaklaşım sunmaktadır. Çalışmada Histogram of Oriented Gradients (HOG), Local Binary Patterns (LBP), Scale-Invariant Feature Transform (SIFT), Gray Level Co-occurrence Matrix (GLCM), Gabor filtreleri, renk histogramı, dalgacık (wavelet) dönüşümü, Hu momentleri, Zernike momentleri ve Fourier dönüşümü olmak üzere on farklı öznitelik çıkarım yöntemi kullanılmıştır. Bu yöntemler, 300×300 piksel boyutundaki gri tonlamalı görüntülere beş farklı hücre boyutunda (25×25, 50×50, 100×100, 150×150 ve 300×300) uygulanmıştır. Yüksek boyutluluk sorununu azaltmak amacıyla minimum Redundancy Maximum Relevance (mRMR) ve Ki-kare (χ²) testleri ile öznitelik seçimi gerçekleştirilmiştir. Elde edilen öznitelik kümeleri, MATLAB Classification Learner aracı kullanılarak Fine Tree, Fine KNN, Wide Neural Network, Bagged Trees, Fine Gaussian SVM ve Binary GLM algoritmalarıyla sınıflandırılmıştır. Deneysel sonuçlara göre en yüksek doğruluk oranı %99,7 ile 25×25 hücre boyutunda tüm özniteliklerin kullanıldığı Wide Neural Network modeliyle elde edilmiştir. Bulgular, küçük hücre boyutlarının yüzey ayrıntılarını daha iyi koruyarak sınıflandırma başarımını artırdığını ortaya koymaktadır. Sonuç olarak önerilen yöntem, döküm endüstrisinde yüksek doğruluklu, hızlı ve uygulanabilir bir kalite kontrol sistemi olarak değerlendirilebilecek potansiyele sahiptir.

Anahtar Kelimeler

• Döküm kusurları
• Görüntü işleme
• Makine öğrenmesi
• Özellik çıkarımı
• HOG
• LBP

A Multi-Feature Extraction and Selection Framework for High-Accuracy Surface Defect Classification in Industrial Casting Parts

Öz

This study presents an automated quality inspection approach for detecting surface defects in industrial casting parts using image processing and machine learning techniques. A total of ten different feature extraction methods—Histogram of Oriented Gradients (HOG), Local Binary Patterns (LBP), Scale-Invariant Feature Transform (SIFT), Gray Level Co-occurrence Matrix (GLCM), Gabor filters, color histogram, wavelet transform, Hu moments, Zernike moments, and Fourier transform—were applied to 300×300 grayscale images. To evaluate the effect of spatial resolution, features were extracted at five different cell sizes: 25×25, 50×50, 100×100, 150×150, and 300×300. Dimensionality reduction was performed using minimum Redundancy Maximum Relevance (mRMR) and Chi-square (χ²) feature selection techniques. The resulting feature sets were classified with six different algorithms in MATLAB Classification Learner, including Fine Tree, Fine KNN, Wide Neural Network, Bagged Trees, Fine Gaussian SVM, and Binary GLM. Experimental results demonstrated that the highest accuracy rate of 99.7% was achieved with the Wide Neural Network model trained on the complete feature set at a 25×25 cell size. These findings indicate that smaller cell sizes preserve critical surface details and enhance classification performance. The study highlights that the proposed methodology can serve as a highly accurate, efficient, and practical solution for automated defect detection in the casting industry, offering strong potential for real-world industrial applications.

Anahtar Kelimeler

• Casting defects
• Image processing
• Machine learning
• Feature extraction
• HOG
• LBP

Kaynakça

1. Aghdam SR, Amid E & Imani MF A fast method of steel surface defect detection using decision trees applied to LBP based features. 2012 7th IEEE Conference on Industrial Electronics and Applications (ICIEA), 2012. IEEE, 1447-1452.
2. Al-Hameed W & Fadel N (2019) Defect detection based optimized threshold of decision tree. Journal of computational and theoretical nanoscience, 16(3):914-919.
3. Al-Rawi M (2023) Ultra-Fast Zernike Moments using FFT and GPU. arXiv preprint arXiv:2304.14492,
4. Andriosopoulou, G., Mastakouris, A., Masouros, D., Benardos, P., Vosniakos, G.-C. & Soudris, D. (2023) Defect Recognition in High-Pressure Die-Casting Parts Using Neural Networks and Transfer Learning. Metals, 13(6):1104.
5. Asha V, Bhajantri NU & Nagabhushan P (2012) Automatic detection of texture-defects using texture-periodicity and Jensen-Shannon divergence. Journal of Information Processing Systems, 8(2):359-374.
6. Belila D, Khaldi B & Aiadi O. (2024) Wavelet Texture Descriptor for Steel Surface Defect Classification. Materials, 17(23):5873.
7. Bilik S & Horak K (2022) SIFT and SURF based feature extraction for the anomaly detection. arXiv preprint arXiv:2203.13068,
8. Chen S & Kaufmann T (2021) Development of data-driven machine learning models for the prediction of casting surface defects. Metals, 12(1):1.

9. Dabhi RR 2020. Real-life industrial dataset of casting product [Online]. Kaggle. Available: https://www.kaggle.com/datasets/ravirajsinh45/real-life-industrial-dataset-of-casting-product [Accessed March 10 2024].
10. Ding C & Peng H (2005) Minimum redundancy feature selection from microarray gene expression data. Journal of bioinformatics and computational biology, 3(02):185-205.
11. Liu M, Fai Cheung C, Senin N, Wang S, Su R & Leach R (2020) On-machine surface defect detection using light scattering and deep learning. Journal of the Optical Society of America A, 37(9):B53-B59.
12. Liu X, Wu L, Guo X, Andriukaitis D, Królczyk G & Li Z (2023) A novel approach for surface defect detection of lithium battery based on improved K-nearest neighbor and Euclidean clustering segmentation. The International Journal of Advanced Manufacturing Technology, 127(1):971-985.
13. Mao K, Wei P, Wang Y, Liu M, Wang S & Zheng N (2025) - CSDD: A Benchmark Dataset for Casting Surface Defect Detection and Segmentation. - IEEE/CAA Journal of Automatica Sinica, - 12(- 5):- 947.
14. Rodríguez JAM (2023) Micro-Scale Surface Recognition via Microscope System Based on Hu Moments Pattern and Micro Laser Line Projection. Metals, 13(5):889.
15. Sun H, Zhou W, Yang J, Shao Y, Zhang L & Mao Z (2025) A Multi-Dimensional Feature Extraction Model Fusing Fractional-Order Fourier Transform and Convolutional Information. Fractal and Fractional, 9(8):533.
16. Szeghalmy S & Fazekas A (2023) A Comparative Study of the Use of Stratified Cross-Validation and Distribution-Balanced Stratified Cross-Validation in Imbalanced Learning. Sensors, 23(4):2333.
17. Thaseen IS & Kumar CA (2017) Intrusion detection model using fusion of chi-square feature selection and multi class SVM. Journal of King Saud University-Computer and Information Sciences, 29(4):462-472.
18. Wang C, Hu J, Yang C & Hu P (2024a) DES-YOLO: a novel model for real-time detection of casting surface defects. PeerJ Comput Sci, 10, e2224.
19. Wang S & Bi X (2025) A Review of Automatic Casting Defect Recognition Methods Based on Deep Learning. International Journal of Mechanical and Electrical Engineering, 7(1):42-49.
20. Wang W, Chen J, Han G, Shi X & Qian G (2024b) Application of Object Detection Algorithms in Non-Destructive Testing of Pressure Equipment: A Review. Sensors, 24(18):5944.
21. Wilimitis D & Walsh CG (2023) Practical Considerations and Applied Examples of Cross-Validation for Model Development and Evaluation in Health Care: Tutorial. Jmir ai, 2, e49023.
22. Yates LA, Aandahl Z, Richards SA & Brook BW (2023) Cross validation for model selection: A review with examples from ecology. Ecological Monographs, 93(1):e1557.
23. Yousef, N., Parmar, C. & Sata, A. (2022) Intelligent inspection of surface defects in metal castings using machine learning. Materials Today: Proceedings, 67(517-522.
24. Zhang, Y., Gao, Z., Sun, J. & Liu, L. (2023) Machine-Learning Algorithms for Process Condition Data-Based Inclusion Prediction in Continuous-Casting Process: A Case Study. Sensors, 23(15):6719.
25. Zhao Z & Wu T (2022) Casting Defect Detection and Classification of Convolutional Neural Network Based on Recursive Attention Model. Scientific Programming, 2022(1):4385565.