Roads are a strategic asset of a country and are of great importance for the movement of passengers and goods. Increasing traffic volume and load, together with the aging of roads, creates various types of anomalies on the road surface. This work proposes a low-cost system for real-time screening of road pavement conditions. Acceleration signals provided by on-car sensors are processed in the time-frequency domain in order to extract information about the condition of the road surface. More specifically, a short-time Fourier transform is used, and significant features, such as the coefficient of variation and the entropy computed over the energy of segments of the signal, are exploited to distinguish between well-localized pavement distresses caused by potholes and manhole covers and spread distress due to fatigue cracking and rutting. The extracted features are fed to supervised machine learning classifiers in order to distinguish the pavement distresses. System performance is assessed using real data, collected by sensors located on the car's dashboard and floorboard and manually labeled. The experimental results show that the proposed system is effective at detecting the presence and the type of distress with high classification rates.

Road Surface Anomaly Assessment Using Low-Cost Accelerometers: A Machine Learning Approach / Martinelli, Alessio; Meocci, Monica; Dolfi, Marco; Branzi, Valentina; Morosi, Simone; Argenti, Fabrizio; Berzi, Lorenzo; Consumi, Tommaso. - In: SENSORS. - ISSN 1424-8220. - STAMPA. - 22:(2022), pp. 3788-3788. [10.3390/s22103788]

Road Surface Anomaly Assessment Using Low-Cost Accelerometers: A Machine Learning Approach

Martinelli, Alessio;Meocci, Monica;Dolfi, Marco;Branzi, Valentina;Morosi, Simone
;
Argenti, Fabrizio;Berzi, Lorenzo;Consumi, Tommaso
2022

Abstract

Roads are a strategic asset of a country and are of great importance for the movement of passengers and goods. Increasing traffic volume and load, together with the aging of roads, creates various types of anomalies on the road surface. This work proposes a low-cost system for real-time screening of road pavement conditions. Acceleration signals provided by on-car sensors are processed in the time-frequency domain in order to extract information about the condition of the road surface. More specifically, a short-time Fourier transform is used, and significant features, such as the coefficient of variation and the entropy computed over the energy of segments of the signal, are exploited to distinguish between well-localized pavement distresses caused by potholes and manhole covers and spread distress due to fatigue cracking and rutting. The extracted features are fed to supervised machine learning classifiers in order to distinguish the pavement distresses. System performance is assessed using real data, collected by sensors located on the car's dashboard and floorboard and manually labeled. The experimental results show that the proposed system is effective at detecting the presence and the type of distress with high classification rates.
22
3788
3788
Martinelli, Alessio; Meocci, Monica; Dolfi, Marco; Branzi, Valentina; Morosi, Simone; Argenti, Fabrizio; Berzi, Lorenzo; Consumi, Tommaso
File in questo prodotto:
File Dimensione Formato  
Articolo pubblicato (early access version 16.05.2022)_redux.pdf

accesso aperto

Descrizione: Early Access Version
Tipologia: Pdf editoriale (Version of record)
Licenza: Open Access
Dimensione 517.49 kB
Formato Adobe PDF
517.49 kB Adobe PDF Visualizza/Apri

I documenti in FLORE sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/2158/1280648
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 1
  • ???jsp.display-item.citation.isi??? 1
social impact