- 112 views
- 132 downloads
Imaging-Based Identification of International Roughness Index Using Deep Neural Networks
-
- Author / Creator
- Zeng, Jiangyu
-
The International Roughness Index (IRI) is one of the most critical indicators in the field of pavement performance management. Since the introduction of Deep Neural Networks (DNNs), many researchers have created different kinds of DNN models to forecast the IRI. However, none of them employed pavement photos as the direct inputs when training their models, regardless of the substantial development of DNNs in image processing. On the other hand, thanks to the fast products in photography equipment, small and convenient sports action cameras, such as the GoPro Hero series, are able to capture smooth videos at a high framerate with built-in electronic image stabilization systems. Although there exist DNN-based models to detect, classify, and quantify distresses as well as models based on neural networks to identify the IRI using distresses, there is currently no model that directly uses pavement images as the input to estimate the IRI. Therefore, this thesis can be considered as the first attempt to use imaging-based DNN models to estimate IRI. Besides the development of the DNN, the procedures of collecting IRI ground truths and stitching a full-size image of an entire 50-m long pavement segment are also demonstrated. After analyzing the testing results, it is shown that imaging-based DNN models can be a promising method to identify the IRI. In the discussion section, several controlled experiments were performed to learn how the related hyperparameters and data combinations affect the performance. Additionally, the proposed DNN model was compared with other models, and the results indicated that the current model still has room for improvement. Therefore, at the end of this thesis, recommendations and future works are also addressed.
-
- Subjects / Keywords
-
- Graduation date
- Spring 2022
-
- Type of Item
- Thesis
-
- Degree
- Master of Science
-
- License
- This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.