Detection of Potholes and Speedbumps by Monitoring Front Traffic

Karthikeyan Nagarajan, Amit Jindal

Research output: Contribution to journalConference article

Abstract

This article proposes a novel method for detecting potholes and speed bumps by monitoring the front traffic when the road is not visibly seen. The existing camera-based systems in the car directly scan the road surface and estimate the road profile. The main disadvantage of the current technology is that it is not possible to detect the road profile when the road is not clearly visible. This can happen in situations when roads are waterlogged or have occluded bumps and in low light conditions. In thisarticle, the proposed method for detecting potholes or speed bumps by monitoring the vertical movement of the front traffic can enhance the existing algorithm by overcoming the abovementioned disadvantage. However, the method works only when there is traffic ahead of the system vehicle. The method makes level 3 and above autonomous driving more robust in terms of comfort and safety by estimating the road profile in the abovementioned road conditions. The method uses an object localization algorithm and optical flow to determine the motion vectors of vehicles. The histogram of motion vectors in the vertical direction is calculated and its weighted average is obtained for each frame. As a feature extraction process, fast Fourier transform (FFT) is applied to the fixed size queue buffer containing weighted average values to obtain the frequency spectrum, which is used as a feature set for classification. After feature selection, the support vector machine (SVM) is used to classify whether a pothole/road bump has been detected. Apart from hardware video stabilization, the same classification procedure is applied to static objects in the frame, and camera egomotion is detected. This is done to avoid false positives due to camera movement when the system vehicle is moving over a pothole or a speed bump. The developed algorithm is validated with the manually collected dataset and the results and the analysis are presented here.

Original languageEnglish
JournalSAE Technical Papers
Volume2019-January
Issue numberJanuary
DOIs
Publication statusPublished - 18-01-2019
Externally publishedYes
Event2019 SAE Automotive Technical Papers, WONLYAUTO 2019 - Warrendale, United States
Duration: 01-01-201901-01-2019

Fingerprint

Cameras
Monitoring
Feature extraction
Optical flows
Fast Fourier transforms
Support vector machines
Railroad cars
Stabilization
Hardware

All Science Journal Classification (ASJC) codes

  • Automotive Engineering
  • Safety, Risk, Reliability and Quality
  • Pollution
  • Industrial and Manufacturing Engineering

Cite this

Nagarajan, Karthikeyan ; Jindal, Amit. / Detection of Potholes and Speedbumps by Monitoring Front Traffic. In: SAE Technical Papers. 2019 ; Vol. 2019-January, No. January.
@article{3773a2fedd5d4f6a9444b4545d93a3ba,
title = "Detection of Potholes and Speedbumps by Monitoring Front Traffic",
abstract = "This article proposes a novel method for detecting potholes and speed bumps by monitoring the front traffic when the road is not visibly seen. The existing camera-based systems in the car directly scan the road surface and estimate the road profile. The main disadvantage of the current technology is that it is not possible to detect the road profile when the road is not clearly visible. This can happen in situations when roads are waterlogged or have occluded bumps and in low light conditions. In thisarticle, the proposed method for detecting potholes or speed bumps by monitoring the vertical movement of the front traffic can enhance the existing algorithm by overcoming the abovementioned disadvantage. However, the method works only when there is traffic ahead of the system vehicle. The method makes level 3 and above autonomous driving more robust in terms of comfort and safety by estimating the road profile in the abovementioned road conditions. The method uses an object localization algorithm and optical flow to determine the motion vectors of vehicles. The histogram of motion vectors in the vertical direction is calculated and its weighted average is obtained for each frame. As a feature extraction process, fast Fourier transform (FFT) is applied to the fixed size queue buffer containing weighted average values to obtain the frequency spectrum, which is used as a feature set for classification. After feature selection, the support vector machine (SVM) is used to classify whether a pothole/road bump has been detected. Apart from hardware video stabilization, the same classification procedure is applied to static objects in the frame, and camera egomotion is detected. This is done to avoid false positives due to camera movement when the system vehicle is moving over a pothole or a speed bump. The developed algorithm is validated with the manually collected dataset and the results and the analysis are presented here.",
author = "Karthikeyan Nagarajan and Amit Jindal",
year = "2019",
month = "1",
day = "18",
doi = "10.4271/2019-01-5031",
language = "English",
volume = "2019-January",
journal = "SAE Technical Papers",
issn = "0148-7191",
publisher = "SAE International",
number = "January",

}

Detection of Potholes and Speedbumps by Monitoring Front Traffic. / Nagarajan, Karthikeyan; Jindal, Amit.

In: SAE Technical Papers, Vol. 2019-January, No. January, 18.01.2019.

Research output: Contribution to journalConference article

TY - JOUR

T1 - Detection of Potholes and Speedbumps by Monitoring Front Traffic

AU - Nagarajan, Karthikeyan

AU - Jindal, Amit

PY - 2019/1/18

Y1 - 2019/1/18

N2 - This article proposes a novel method for detecting potholes and speed bumps by monitoring the front traffic when the road is not visibly seen. The existing camera-based systems in the car directly scan the road surface and estimate the road profile. The main disadvantage of the current technology is that it is not possible to detect the road profile when the road is not clearly visible. This can happen in situations when roads are waterlogged or have occluded bumps and in low light conditions. In thisarticle, the proposed method for detecting potholes or speed bumps by monitoring the vertical movement of the front traffic can enhance the existing algorithm by overcoming the abovementioned disadvantage. However, the method works only when there is traffic ahead of the system vehicle. The method makes level 3 and above autonomous driving more robust in terms of comfort and safety by estimating the road profile in the abovementioned road conditions. The method uses an object localization algorithm and optical flow to determine the motion vectors of vehicles. The histogram of motion vectors in the vertical direction is calculated and its weighted average is obtained for each frame. As a feature extraction process, fast Fourier transform (FFT) is applied to the fixed size queue buffer containing weighted average values to obtain the frequency spectrum, which is used as a feature set for classification. After feature selection, the support vector machine (SVM) is used to classify whether a pothole/road bump has been detected. Apart from hardware video stabilization, the same classification procedure is applied to static objects in the frame, and camera egomotion is detected. This is done to avoid false positives due to camera movement when the system vehicle is moving over a pothole or a speed bump. The developed algorithm is validated with the manually collected dataset and the results and the analysis are presented here.

AB - This article proposes a novel method for detecting potholes and speed bumps by monitoring the front traffic when the road is not visibly seen. The existing camera-based systems in the car directly scan the road surface and estimate the road profile. The main disadvantage of the current technology is that it is not possible to detect the road profile when the road is not clearly visible. This can happen in situations when roads are waterlogged or have occluded bumps and in low light conditions. In thisarticle, the proposed method for detecting potholes or speed bumps by monitoring the vertical movement of the front traffic can enhance the existing algorithm by overcoming the abovementioned disadvantage. However, the method works only when there is traffic ahead of the system vehicle. The method makes level 3 and above autonomous driving more robust in terms of comfort and safety by estimating the road profile in the abovementioned road conditions. The method uses an object localization algorithm and optical flow to determine the motion vectors of vehicles. The histogram of motion vectors in the vertical direction is calculated and its weighted average is obtained for each frame. As a feature extraction process, fast Fourier transform (FFT) is applied to the fixed size queue buffer containing weighted average values to obtain the frequency spectrum, which is used as a feature set for classification. After feature selection, the support vector machine (SVM) is used to classify whether a pothole/road bump has been detected. Apart from hardware video stabilization, the same classification procedure is applied to static objects in the frame, and camera egomotion is detected. This is done to avoid false positives due to camera movement when the system vehicle is moving over a pothole or a speed bump. The developed algorithm is validated with the manually collected dataset and the results and the analysis are presented here.

UR - http://www.scopus.com/inward/record.url?scp=85067821281&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85067821281&partnerID=8YFLogxK

U2 - 10.4271/2019-01-5031

DO - 10.4271/2019-01-5031

M3 - Conference article

VL - 2019-January

JO - SAE Technical Papers

JF - SAE Technical Papers

SN - 0148-7191

IS - January

ER -