國立虎尾科技大學 |

Detecting Potholes Using Deep Neural Networks with Unmanned Aerial Vehicles.

紀錄類型:	書目-語言資料,手稿 : Monograph/item
正題名/作者:	Detecting Potholes Using Deep Neural Networks with Unmanned Aerial Vehicles./
作者:	Welborn, Ethan Andrew.
面頁冊數:	1 online resource (44 pages)
附註:	Source: Masters Abstracts International, Volume: 84-11.
Contained By:	Masters Abstracts International84-11.
標題:	Computer engineering. -
電子資源:	click for full text (PQDT)
ISBN:	9798379554149

Detecting Potholes Using Deep Neural Networks with Unmanned Aerial Vehicles.
Welborn, Ethan Andrew.

Detecting Potholes Using Deep Neural Networks with Unmanned Aerial Vehicles. - 1 online resource (44 pages)

Source: Masters Abstracts International, Volume: 84-11.

Thesis (M.S.)--Tarleton State University, 2023.

Includes bibliographical references

Potholes are a problem for both urban and rural communities, and local governments are stretched thin maintaining their roads. Locating potholes is the number one step to fixing them. There are several methods to detecting potholes, including sensors, 3D visualization with lasers (LiDAR), or simply by visual inspection of the road, but considering recent innovations in the field of computer vision with deep neural networks, or DNNs, being used for object detection and instance segmentation and achieving fast and accurate results, and the availability of high quality unmanned aerial vehicles, UAVs, or drones, to the civilian market, it is economically feasible and reasonable for DNNs and UAVs to be combined for pothole detection. This work explores the approach of applying UAVs and DNNs to the pothole detection problem, in order to find a more efficient and robust solution to detecting potholes.Three DNN architectures were reviewed, Mask R-CNN, YOLO X, and YOLOv8, to determine which architecture is better suited for pothole detection. They were separately trained on both a ground-level dataset, containing images of potholes taken from ground level with a phone or handheld camera, and a mixed dataset, containing ground level images along with aerial images taken with UAVs or from the top of ladders - the results from these two datasets were compared to determine which angle and altitude is preferred for detecting potholes, and if training on aerial images is necessary for a DNN to successfully detect potholes from UAV images.The most accurate model, YOLOv8, achieved a MAP50 of 0.630 for detecting potholes and 0.643 for segmenting, or determining the shape of, the potholes. YOLOv8 was also the quickest model, achieving an inference speed of 3.6 milliseconds, or 278 frames per second.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2024

Mode of access: World Wide Web

ISBN: 9798379554149Subjects--Topical Terms:

569006
Computer engineering.
Subjects--Index Terms:

Computer visionIndex Terms--Genre/Form:

554714
Electronic books.

Detecting Potholes Using Deep Neural Networks with Unmanned Aerial Vehicles.
LDR:03157ntm a22003977 4500 001 1148152
005 20240916070014.5
006 m o d
007 cr bn ---uuuuu
008 250605s2023 xx obm 000 0 eng d
020 $a 9798379554149
035 $a (MiAaPQ)AAI30418756
035 $a AAI30418756
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Welborn, Ethan Andrew. $3 1474063
245 1 0 $a Detecting Potholes Using Deep Neural Networks with Unmanned Aerial Vehicles.
264 0 $c 2023
300 $a 1 online resource (44 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Masters Abstracts International, Volume: 84-11.
500 $a Advisor: Diamantas, Sotirios.
502 $a Thesis (M.S.)--Tarleton State University, 2023.
504 $a Includes bibliographical references
520 $a Potholes are a problem for both urban and rural communities, and local governments are stretched thin maintaining their roads. Locating potholes is the number one step to fixing them. There are several methods to detecting potholes, including sensors, 3D visualization with lasers (LiDAR), or simply by visual inspection of the road, but considering recent innovations in the field of computer vision with deep neural networks, or DNNs, being used for object detection and instance segmentation and achieving fast and accurate results, and the availability of high quality unmanned aerial vehicles, UAVs, or drones, to the civilian market, it is economically feasible and reasonable for DNNs and UAVs to be combined for pothole detection. This work explores the approach of applying UAVs and DNNs to the pothole detection problem, in order to find a more efficient and robust solution to detecting potholes.Three DNN architectures were reviewed, Mask R-CNN, YOLO X, and YOLOv8, to determine which architecture is better suited for pothole detection. They were separately trained on both a ground-level dataset, containing images of potholes taken from ground level with a phone or handheld camera, and a mixed dataset, containing ground level images along with aerial images taken with UAVs or from the top of ladders - the results from these two datasets were compared to determine which angle and altitude is preferred for detecting potholes, and if training on aerial images is necessary for a DNN to successfully detect potholes from UAV images.The most accurate model, YOLOv8, achieved a MAP50 of 0.630 for detecting potholes and 0.643 for segmenting, or determining the shape of, the potholes. YOLOv8 was also the quickest model, achieving an inference speed of 3.6 milliseconds, or 278 frames per second.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2024
538 $a Mode of access: World Wide Web
650 4 $a Computer engineering. $3 569006
650 4 $a Computer science. $3 573171
653 $a Computer vision
653 $a Deep learning
653 $a Deep neural networks
653 $a Object detection
653 $a Pothole
655 7 $a Electronic books. $2 local $3 554714
690 $a 0984
690 $a 0464
690 $a 0800
710 2 $a Tarleton State University. $b Department of Engineering and Computer Science. $3 1474064
710 2 $a ProQuest Information and Learning Co. $3 1178819
773 0 $t Masters Abstracts International $g 84-11.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30418756 $z click for full text (PQDT)