Hybrid Model for Detection of Corrosion in Water Pipeline Images Using CNN and Comparing Accuracy with SVM

Abstract

The work aims at studying a hybrid model for novel corrosion detection in water pipeline images using two different machine learning algorithms in low resolution images. Methods and Material: Convolutional Neural Network (CNN) and Support Vector Machine (SVM) algorithm implemented to detect the corrosion in low resolution image dataset with 40 samples. Results: CNN Classifier model has an detection accuracy value of 93.18% and the SVM has an detection accuracy of 77.77%. Attained significance (p=0.001) through SPSS tool. Conclusion: CNN algorithm perform well compared to SVM algorithm.