Low-cost deep learning UAV and Raspberry Pi solution to real time pavement condition assessment

In this thesis, a real-time and low-cost solution to the autonomous condition assessment of pavement is proposed using deep learning, Unmanned Aerial Vehicle (UAV) and Raspberry Pi tiny computer technologies, which makes roads maintenance and renovation management more efficient and cost effective. A comparison study was conducted to compare the performance of seven different combinations of meta-architectures for pavement distress classification. It was observed that real-time object detection architecture SSD with MobileNet feature extractor is the best combination for real-time defect detection to be used by tiny computers. A low-cost Raspberry Pi smart defect detector camera was configured using the trained SSD MobileNet v1, which can be deployed with UAV for real-time and remote pavement condition assessment. The preliminary results show that the smart pavement detector camera achieves an accuracy of 60% at 1.2 frames per second in raspberry pi and 96% at 13.8 frames per second in CPU-based computer.

Grantor dc:publisher

University of Tennessee at Chattanooga

Author dc:creator

• Qurishee, Murad Al

Contributors dc:contributor

• Wu, Weidong; Owino, Joseph
• Fomunung, Ignatius; Onyango, Mbakisya A.; Liang, Yu
• College of Engineering and Computer Science

• Pavements -- Design and construction

Licence

• https://rightsstatements.org/page/InC/1.0/?language=en

Language dc:language

English, eng

Repository record dc:identifier

https://scholar.utc.edu/theses/601

OAI identifier oai:identifier

oai:scholar.utc.edu:theses-1752