Application of Histogram of Oriented Gradients and Support Vector Machine on Detection of Far-side Corrosion

Abstract

Nondestructive testing (NDT) of far-side corrosion in a multilayer structure of aircraft is an important task to ensure the integrity and safety of the aircraft. Among the NDT methods, electromagnetic testing (ET) is powerful in detecting far-side corrosion. However, the far-side corrosion usually appears at the riveting site, making the ET signal complicated and challenging to recognize the presence of the small corrosion. In this paper, a histogram of gradients will be applied to extract features of the magnetic image, and a support vector machine will be used to detect the far-side corrosion. The proposed method helps to improve the accuracy of the detection significantly.