Qualitative comparison of methodologies for detecting surface defects in aircraft interiors

Automated identification of parts showing defects have established itself as one of the key aspects in production factories during the second and third industrial revolutions, especially in sectors, such as the automotive, enjoying economies of scale. Automation can be exploited even in successive steps too, such as the final assembly of large manufactured goods, like aircrafts, where the usage of automated systems can provide judgement consistencies that may be challenging to obtain in more traditional ways. In this paper, different methodologies for the identification of visual features are compared together. These approaches cover both traditional computer vision techniques, such as SURF descriptors and machine learning based algorithms, namely a convolutional neural network, for the identification of surface defects in aircraft interiors. The comparison is performed at a qualitative level, enabling a discussion between pros and cons of each approach.