Detection of missing nuts & bolts on rail fishplate

In India human operator manually inspects the fishplates on railroads for any missing nut & bolt; broadly these inspections are based on visual analysis that might be prone to human errors, endangering passengers of the train. This paper is based on a machine vision approach that emulates the visual ability of human operator for the automation of the process. Digital im ages have been acquired of rail fishplates from both sides (left and right) of rail joint. A pattern recognition approach has been adopted for detecting the existence of nuts & bolts of fishplate at rail joints and thereby achieves automation. The number of nuts & bolts on fishplate, length, and width of the fishplate computed as features through discrete wavelet transform (DWT). The features evaluated relative to the database features and inference estimated in terms of the fitness of the fishplate. The effect of using DWT along with the similarity measure has been tested through performance metrics. Results obtained shows that the proposed approach has high reliability and robustness with less computational complexity.