Detection of the Road Pothole Contour in Raster Images

Abstract

The article analyses the issues related to detection and contour approximation of road surface defects, in particular, road potholes captured by a smart device camera. The assessment of measurements, obtained after image analysis and real objects, will be disclosed with a view to depict the measurement bias. The difference from all the other methods, mentioned in the literature, is that only 2D images captured by one camera were used in this investigation with a view to identify the contour of road potholes and get its parameterised representation. In practice, according to other authors, the most common approach to obtain a parameterised description of a pothole incorporates not only cameras, but also other various sensors such as accelerometer, global positioning systems, laser, hyperspectral imagery, infrared or ultrasonic sensors. In this article, we present a method that allows recognising a pothole object in terms of its colour, shape, and structure. The method discussed was applied to real world images to detect and outline the road pothole contour. Finally, the evaluation of approximation accuracy by empirical research techniques has been accomplished.

Automatic Recognition of Road Damage Based on Lightweight Attentional Convolutional Neural Network

“…Besides geometric morphology, some studies perform road damage detection from the perspective of image colour. For example, Jakštys et al [23] outlined the edge contours of road potholes by analysing the B-component in the RGB colour space of road potholes in the ROI region. Akagic et al [24], on the other hand, analysed the B-component in the RGB colour space by component to perform image segmentation of the asphalt pavement and then detected the pothole areas by processing methods such as cropping, Otsu thresholding, and boundary elimination.…”

Detecção de anomalias em pavimentos utilizando sensores inerciais.

“…No processamento clássico técnicas para extrair informações da matriz de cor RGB (Red,Green,Blue) são aplicadas, normalmente na forma de um processamento sequencial [60] combinando diversas análises como filtro de cores, agrupamento espectral [61], limiar de constraste [62], extração de textura [63] etc.…”