Contents lists available at ScienceDirect Measurement journal homepage: www.elsevier.com/locate/measurement Digital image analysis technique for measuring railway track defects and ballast gradation Marco Guerrieri a , Giuseppe Parla b , Clara Celauro b, ⁎ a DICAM – Department of Civil, Environmental and Mechanical Engineering, University of Trento, Italy b DICAM – Dipartimento di Ingegneria Civile, Ambientale, Aerospaziale, dei Materiali, Università degli Studi di Palermo, Italy ARTICLE INFO Keywords: Railway track Rail profile Ballast Aggregate gradation Image analysis Stereoscopic vision ABSTRACT In order to guarantee safety and driving comfort and to maintain an efficient railway infrastructure, the first step is to carefully monitor the track geometry and wear level of the materials constituting the superstructure. To that end diagnostic trains are widely used on main lines, in that they can detect several geometric track parameters and rail wear, but under no circumstances they can yet detect ballast gradation. Due to the practical implications for the planning of maintenance operations on the railway network, this article presents a “DIP” digital image processing technique for measuring the transverse profile and corrugations of the rails as well as ballast gradation. The research was carried out in the laboratory on samples of worn-out rails taken from operational railway lines and in situ in the case of ballast analyses. For the latter, the reliability of the results obtained was assessed by comparison with available results yielded by traditional testing methods. It is shown that the proposed technique can be used not only for laboratory analyses, but most conveniently for high-efficiency in situ surveys, along with the methods traditionally adopted by the railway managing autho- rities thus contributing to lowering the maintenance cost associated with rail inspection. 1. Introduction Railway track wear is a primary source of both user discomfort and safety problems. Above all, the rail-head wear can be a concomitant cause of derailment following train wheels climbing on the track. This case occurs when the forward motion of the axle combines with an excessive ratio of Q/P (wheel/rail contact forces), usually just when reduced vertical force and increased lateral force act on the wheel flange so that it rolls onto the top of the rail head. The climb condition may be temporary, with wheel and rail returning to normal contact, or it may result in the wheel climbing fully over the rail [1]. By denoting the transverse load, the wheel load, the flange angle and the friction coefficient with Q, P, α and f respectively, the limit value of the ratio Q/ P which avoids derailment can be obtained with Nadal’s well-known formula [2]: = − + Q P tan(α) f 1 ftan(α) (1) Relation (1) shows that when the flange angle increases, the ratio Q/P decreases, thus maximizing the derailment risk. This circumstance arises as an effect of rail wear (α increment). On the other hand, the different types of irregularity of the rail rolling surface and especially of the corrugation due to the wheel/rail parametric excitation [2–4] lead to N increase in dynamic load [5] to which the track is subjected, with a consequent rapid functional decay of the railway superstructure (so called “railway track”) and an increase in rolling noise (prevailing over the other railway noise sources in the speed range 40–200 km/h). Rail reclamation and removal of transverse and longitudinal wear faults are achieved with “rail grinding machines” (e.g. the Plasser GWM 250 rail-grinding machine [5]). In order to keep the infrastructure efficient, reduce derailment risk and limit noise emission, maintenance activities need to be adequately planned after proper monitoring of the railway superstructure and identification of primary distresses in track components [6,7]. The same happens for road infrastructures or in civil concrete structures and image processing is becoming an extremely useful tool for this kind of applications [8–10]. Distress detection and monitoring for railways (mainly focusing on rail profile and level as well as overall geometry and ondulation) is typically based on mechanical devices in contact with the track, or via innovative approaches based on laser scanning and image analysis [11–13]. Amongst the contact methods it is worth mentioning: http://dx.doi.org/10.1016/j.measurement.2017.08.040 Received 9 December 2015; Received in revised form 9 June 2017; Accepted 28 August 2017 ⁎ Corresponding author. E-mail addresses: marco.guerrieri@unipa.it (M. Guerrieri), giuseppe.parla@unipa.it (G. Parla), clara.celauro@unipa.it (C. Celauro). Measurement 113 (2018) 137–147 0263-2241/ © 2017 Elsevier Ltd. All rights reserved. MARK