Automated groove identification and measurement using long short-term memory unit

Transverse grooves have been widely used on airport runways to improve their drainage capacity and skid resistance. Therefore, regular measurement and evaluation of groove dimensions on airport runways are significant for runways to improve skid resistance and eliminate hydroplaning risks. However, there are few effective methods that are able to automatically and accurately measure groove dimension. This study introduces a new method for automated groove identification and measurement using Long Short-term Memory (LSTM). The GrooveNet is designed to identify the potential dip and determine two endpoints of the identified dip. Subsequently, dip dimension is calculated according to FAA AC No. 150/5320-12C. Finally, the modified Naive Bayes classifier is proposed to distinguish grooves and slab joints. Results indicate that the proposed methodology (including GrooveNet and modified Naive Bayes classifier) is more robust and accurate in runway groove identification and measurement. With the proposed approach, operators of airfield runway pavements have a robust tool to conduct groove safety evaluation and further provide corrective maintenance actions for the unsafe runway grooves. (C) 2019 Elsevier Ltd. All rights reserved.