Photogrammetric low-cost unmanned aerial vehicle for pothole detection mapping / Shahrul Nizan Abd Mukti
Types of pavement distress include potholes, cracks and rutting. The most dangerous and common form of road defect is the pothole. Effective data collection of pothole geometry information can assist in road maintenance project. It is necessary in estimating a logistic need which can help Authoritie...
Saved in:
| 主要作者: | |
|---|---|
| 格式: | Thesis |
| 语言: | English |
| 出版: |
2022
|
| 主题: | |
| 在线阅读: | https://ir.uitm.edu.my/id/eprint/76663/1/76663.pdf |
| 标签: |
添加标签
没有标签, 成为第一个标记此记录!
|
| 总结: | Types of pavement distress include potholes, cracks and rutting. The most dangerous and common form of road defect is the pothole. Effective data collection of pothole geometry information can assist in road maintenance project. It is necessary in estimating a logistic need which can help Authorities in distribute their budget projection. Pavement damage involving spectral features has limitations with the RGB format. In the field of remote sensing, road distress in the form of spectral features has been characterised by Multispectral (MS) images that have coverage with broad wavelengths. This study extracted pothole area and volume information from fusion of Digital Elevation Model (DEM) and classified MS image. The study set four main objectives to achieve its aim: (1) To analyse RGB and multispectral sensor calibration, (2) To evaluate the optimal flight parameters for pothole modelling production using RGB imagery, (3) To investigate various classifier algorithms and band combinations for pothole region areas using multispectral imagery and (4) To validate geometric information from the extracted pothole. Three (3) chequerboards were used in camera calibration to find an optimal camera parameter. 50mm size of chequerboard pattern yielded the lowest pre-projection error while there’s no guaranteed in having better projection by increased the focal length of the camera. Meanwhile, nine (9) classifier algorithms and forty (40) band stack combinations were deployed to classify the pothole edge. The most significant classifier algorithms to distinguish a pothole defect is Maximum likelihood with 29 over 40 band combination win rate. The lowest error of pothole polygon classification is 0.016m² from classification of Mahalanobis distance algorithm with NIR + red edge + red or Green + red edge + red combination. For volume estimation purpose, a Digital Elevation Models (DEM) were generated from photogrammetric method with flight parameters of three (3) different camera focal length and ten (10) UAV altitude. The optimal flight parameters for accurate fill-in volume estimation is at 8 m and 10 m flight altitude, with the 3.61 mm and 8.8 mm of focal lengths, respectively. These results confirmed that the UAV is a very useful form of technology in road maintenance applications, at the same time, contributing in alternative method of pothole information extraction. The information gathered is useful for Authorities, concession expressway company and interest party for better road maintenance implementation. |
|---|