Loading...

Crack Detection on Asphalt Concrete Pavements Using Image Processing

Mohammadi, Mohsen | 2022

164 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 55263 (09)
  4. University: Sharif University of Technology
  5. Department: Civil Engineering
  6. Advisor(s): Sabouri, Mohammad Reza
  7. Abstract:
  8. Road transportation has long been an important form of transportation for mankind. Mass production of motorized vehicles and a growing number of people using human and goods transportation have led engineers to monitor road pavement condition and to act on accordingly in the case of degradation. Due to road networks expansion and budget restrictions, road pavement maintenance and rehabilitation projects should be prioritized based on up-to-date data. Automatic collection of data related to the pavement condition such as the number and the length of cracks, could save time and reduce cost and workforce. This research proposes a hybrid method composed of two independent techniques for automatic pavement crack detection and segmentation. The hybrid method initially analyzes asphalt pavement images by utilizing learning-based and density-based techniques separately. The learning-based technique extracts features related to the texture of the image and trains a model by Random Forest. The density-based technique utilizes thresholding as the initial segmentation to create a density map indicating possible crack objects. Various pre-processing and post-processing algorithms were also employed in both techniques. The results are then combined to predict the final crack objects. A prominent property of cracks included in pavement condition surveys is their lengths. A new method has been proposed in this research to estimate the length of linear cracks based on asphalt pavement images. This method comprises crack detection, digital length calculation, and actual length estimation. Several experiments were conducted to evaluate the proposed methods. The hybrid crack segmentation method reached an average precision score of 90%, recall score of 78%, and F1 score of 84%. The crack length estimation method resulted in an average error of less than 3% (2.6020%). These experiments suggest that the proposed methods perform well
  9. Keywords:
  10. Crack Detection ; Pavement Distress ; Image Processing ; Asphalt Pavement ; Crack Length ; Crack Segmentation

 Digital Object List

 Bookmark

No TOC