Road drainage system localisation and condition data capture

Abstract

Inspection of road networks is time and cost consuming. Over half of the money allocated for road maintenance are spent on this area. Visual inspection is still the most commonly employed way of inspection for most parts of the road network. The sheer amount of lane miles of the road network renders this type of road monitoring a costly process. The ultimate goal of this paper is to reduce the time and cost needed to perform routine drive-by, visual and conditional data capturing. The road inspector devotes effort to visually capture the great number of the road assets as it consists of capturing of both the geometry and condition of multiple road assets. In this paper, we propose a method to capture road drainage covers through images and assess if they could potentially be blocked or not. This proposed novel framework focuses on two main tasks: a) localisation of road drainage b) assessment of drainage condition (if they are clean or not). This solution uses Speeded up Robust Features (SURF) and Scale Invariant Feature Form (SIFT) detector, the Bag of Visual Words (BoVW) enhanced with k-means algorithms for grouping the features, and a Support Vector Machine classifier for classifying the data to their respective categories.