Full length article Wavelet-morphology based detection of incipient linear cracks in asphalt pavements from RGB camera imagery and classification using circular Radon transform Yashon O. Ouma a,⇑ , Michael Hahn b a Department of Civil and Structural Engineering, Moi University, Eldoret, Kenya b Department of Geomatics, University of Applied Sciences Stuttgart, Schellingstr, 70174 Stuttgart, Germany article info Article history: Received 3 March 2016 Received in revised form 13 May 2016 Accepted 16 June 2016 Keywords: Micro-linear asphalt pavement distresses Partial area effect (PAE) Discrete wavelet transform (DWT) Successive morphologic transform filtering (SMF) Circular Radon transform (CRT) Dice similarity coefficient abstract The combined detection, extraction and identification of incipient or micro-linear distresses in asphalt pavements are important steps in the quantification and analyses of the occurrences of linear distresses for early pavement management and repair (M&R). This study presents an empirical approach for the for- malized identification of incipient linear structural failures in asphalt pavements, which are characterized by longitudinal, transverse, diagonal, block (random) and alligator (fatigue) distresses. Because of the spectral and spatial complexities in detecting distress features at very high resolutions, this study pre- sents a triple-transform approach for distress detection, isolation and classification that comprises of: (i) 2D discrete wavelet transform (DWT) for multidirectional and multiscale linear distress detection; (ii) successive morphologic transformation filtering (SMF) as an adaptive filter for the extraction of linear distress shape and continuity, and (iii) circular Radon Transform (CRT) for angular-geometric orientation analysis for the identification and classification of the distress types. Using mobile RGB camera imaging, 72 pavement distress images, at a spatial resolution of about 1 mm were selected for evaluating the pro- posed approach. The results of the DWT-SMF were validated using the Dice coefficient of similarity between the manually segmented distresses and the study results. The validation results show that the linear distresses are satisfactorily extracted with an average detection rate of 83.2%. The average pro- cessing time for implementing the DWT-SMF phase of the algorithm was approximately 125 s. To vali- date the classifications of the distress types, the CRT results were matched with the reference classifications from synthetic cracks, with all showing positively corresponding results. In overall, the results of the study illustrate that the proposed triple-transform approach provides a reliable approach for the detection, isolation and characterization of linear distresses in flexible asphalt pavements. Ó 2016 Elsevier Ltd. All rights reserved. 1. Introduction Flexible asphalt pavements are continuously subjected to non- uniform traffic loading, stress, and variations in climatic and envi- ronmental conditions. This, over time results into structural and functional deformations which are manifested by pavement cracks. Cracks or distresses on pavement surfaces reduce the design-life and the economics of the road surface. Through early road condi- tion surveys and analyses, asphalt pavements with incipient dete- riorations can be identified and if maintenance and repair (M&R) is carried out in time, then the overall M&R costs will be reduced by up to 80% [13]. As summarized in Table 1, asphalt pavement failures can be characterized as linear based, area-depth based and surface disin- tegrations as a result of pavement strength failure. The focus in this study is on the detection and extraction of linear distresses, at early or incipient stages, and their characterizations using imagery acquired from mobile RGB camera imaging system. Multichannel RGB camera has the advantage of spectrally discriminating and imaging particle analysis, which enables the separation and char- acterization of the different particle types on a heterogeneous sur- face. In this study, incipient linear distresses are also referred to as micro-linear pavement distresses, and are defined as thin curvilin- ear distresses with widths of not more than 10 mm. As compared to micro-linear distresses, macro-linear distresses are much easier to detect and isolate on paved road surfaces because of their sizes. Typically, asphalt pavement surfaces tend to be very heteroge- neous due to the characteristics of the pavement surface materials http://dx.doi.org/10.1016/j.aei.2016.06.003 1474-0346/Ó 2016 Elsevier Ltd. All rights reserved. ⇑ Corresponding author. E-mail addresses: yashon_o@hotmail.com (Y.O. Ouma), michael.hahn@hft- stuttgart.de (M. Hahn). Advanced Engineering Informatics 30 (2016) 481–499 Contents lists available at ScienceDirect Advanced Engineering Informatics journal homepage: www.elsevier.com/locate/aei