Volume 11, Issue 22 (11-2023)
Ecol Iran For 2023, 11(22): 120-131 |
Back to browse issues page
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
rezaei mutlagh A, parsakhoo A, najafi A, mohamadi J. (2023). Specification of Forest Road Surface Potholes using UAVs Image Evaluation. Ecol Iran For. 11(22), 120-131. doi:10.61186/ifej.11.22.110
URL: http://ifej.sanru.ac.ir/article-1-480-en.html
URL: http://ifej.sanru.ac.ir/article-1-480-en.html
1- Faculty of Natural Resources, Gorgan University of Agricultural Sciences and Natural Resources, Gorgan, Iran
2- Department of Forestry, Faculty of Forest Resources, Gorgan University of Agricultural Sciences and Natural Resources, Gorgan, Iran
3- Department of Forest Engineering, Faculty of Natural Resources and Marine Sciences, TMU, Noor, Iran
2- Department of Forestry, Faculty of Forest Resources, Gorgan University of Agricultural Sciences and Natural Resources, Gorgan, Iran
3- Department of Forest Engineering, Faculty of Natural Resources and Marine Sciences, TMU, Noor, Iran
Abstract: (1223 Views)
Extended Abstract
Background: The surface layer of roads is subject to various forms of damage over time, primarily due to the combined effects of traffic and atmospheric factors. This degradation leads to a significant decline in road quality, which can pose safety risks for users. Therefore, the quick and accurate detection of surface layer anomalies is essential for effective monitoring of road health status. Timely identification of these issues allows for appropriate maintenance and repairs, ultimately enhancing road safety and longevity. To improve the efficiency of surface layer inspections, drones have emerged as valuable tools, providing reliable information in the assessment of road conditions. By utilizing drone technology, it becomes possible to conduct comprehensive evaluations of road surfaces, enabling the identification of specific problems such as potholes, cracks, and other forms of wear.
Methods: This research was conducted with the objective of revealing the top layer damage on forest roads by employing UAV (Unmanned Aerial Vehicle) images and advanced image processing techniques. The study focused on a 3.6 km long road located within the Dr. Bahramnia Educational and Research Forest in Golestan Province. High-resolution images obtained from the UAV were processed using photogrammetric techniques to create orthomosaic images and a digital elevation model of the ground. These processed images served as a foundation for identifying and assessing surface layer anomalies, particularly potholes. To achieve this, three supervised learning algorithms were employed: Nearest Neighbor, K-Nearest Neighbor, and Support Vector Machine (SVM). Each algorithm was rigorously evaluated for its effectiveness in classifying and identifying the various types of damage present in the road surface. The process began with the collection of UAV imagery, which was then subjected to photogrammetric processing. This involved calculations to generate orthomosaic images that accurately represented the road surface. The digital elevation model provided additional context regarding the topography of the area, allowing for a more detailed analysis of the surface layer. Once the images were prepared, they were analyzed using the selected machine learning algorithms. The Nearest Neighbor algorithm assesses the proximity of data points to classify anomalies, while the K-Nearest Neighbor algorithm extends this concept by considering multiple neighboring points for classification. The Support Vector Machine algorithm, on the other hand, uses hyperplanes to distinguish between different classes of data, making it particularly effective for complex datasets.
Results: The results of the study demonstrated that the orthomosaic images generated from photogrammetric calculations exhibited high accuracy, providing a reliable basis for further analysis. The assessment of the algorithms used for pothole classification revealed that all three algorithms displayed strong capabilities in identifying road surface damages. Specifically, the Nearest Neighbor algorithm achieved an accuracy of 92.04%, indicating its effectiveness in recognizing surface anomalies. The K-Nearest Neighbor algorithm performed slightly better, reaching an accuracy of 94.31%. However, the Support Vector Machine algorithm significantly outperformed the others, achieving an impressive accuracy of 96.59%. This high level of accuracy underscores the potential of SVM for effectively classifying and identifying road surface failures.
The successful application of these algorithms not only highlights their individual strengths but also emphasizes the importance of utilizing advanced technologies in road maintenance. The ability to accurately detect and classify potholes can lead to more efficient repair processes and better allocation of resources for road maintenance. Additionally, the integration of UAV imagery with machine learning techniques provides a scalable solution that can be applied to various road types and conditions.
Conclusion: In conclusion, the Support Vector Machine algorithm emerged as the most effective supervised learning method in this study, achieving the highest classification accuracy at 96.59%. This finding suggests that SVM is particularly well-suited for identifying and classifying failures in road surfaces, making it a valuable tool for transportation agencies and road maintenance professionals. The research underscores the potential of UAV imagery combined with machine learning algorithms to enhance the detection of abnormalities in forest roads, including potholes. By adopting these innovative technologies, road management can become more proactive, ensuring safer and more reliable transportation infrastructure in forested areas. Future studies may explore the application of these methods in diverse environments and road conditions, further validating their effectiveness and adaptability in real-world scenarios. The integration of such advanced methodologies can significantly contribute to the sustainability and safety of road networks, ultimately benefiting both users and the environment.
Methods: This research was conducted with the objective of revealing the top layer damage on forest roads by employing UAV (Unmanned Aerial Vehicle) images and advanced image processing techniques. The study focused on a 3.6 km long road located within the Dr. Bahramnia Educational and Research Forest in Golestan Province. High-resolution images obtained from the UAV were processed using photogrammetric techniques to create orthomosaic images and a digital elevation model of the ground. These processed images served as a foundation for identifying and assessing surface layer anomalies, particularly potholes. To achieve this, three supervised learning algorithms were employed: Nearest Neighbor, K-Nearest Neighbor, and Support Vector Machine (SVM). Each algorithm was rigorously evaluated for its effectiveness in classifying and identifying the various types of damage present in the road surface. The process began with the collection of UAV imagery, which was then subjected to photogrammetric processing. This involved calculations to generate orthomosaic images that accurately represented the road surface. The digital elevation model provided additional context regarding the topography of the area, allowing for a more detailed analysis of the surface layer. Once the images were prepared, they were analyzed using the selected machine learning algorithms. The Nearest Neighbor algorithm assesses the proximity of data points to classify anomalies, while the K-Nearest Neighbor algorithm extends this concept by considering multiple neighboring points for classification. The Support Vector Machine algorithm, on the other hand, uses hyperplanes to distinguish between different classes of data, making it particularly effective for complex datasets.
Results: The results of the study demonstrated that the orthomosaic images generated from photogrammetric calculations exhibited high accuracy, providing a reliable basis for further analysis. The assessment of the algorithms used for pothole classification revealed that all three algorithms displayed strong capabilities in identifying road surface damages. Specifically, the Nearest Neighbor algorithm achieved an accuracy of 92.04%, indicating its effectiveness in recognizing surface anomalies. The K-Nearest Neighbor algorithm performed slightly better, reaching an accuracy of 94.31%. However, the Support Vector Machine algorithm significantly outperformed the others, achieving an impressive accuracy of 96.59%. This high level of accuracy underscores the potential of SVM for effectively classifying and identifying road surface failures.
The successful application of these algorithms not only highlights their individual strengths but also emphasizes the importance of utilizing advanced technologies in road maintenance. The ability to accurately detect and classify potholes can lead to more efficient repair processes and better allocation of resources for road maintenance. Additionally, the integration of UAV imagery with machine learning techniques provides a scalable solution that can be applied to various road types and conditions.
Conclusion: In conclusion, the Support Vector Machine algorithm emerged as the most effective supervised learning method in this study, achieving the highest classification accuracy at 96.59%. This finding suggests that SVM is particularly well-suited for identifying and classifying failures in road surfaces, making it a valuable tool for transportation agencies and road maintenance professionals. The research underscores the potential of UAV imagery combined with machine learning algorithms to enhance the detection of abnormalities in forest roads, including potholes. By adopting these innovative technologies, road management can become more proactive, ensuring safer and more reliable transportation infrastructure in forested areas. Future studies may explore the application of these methods in diverse environments and road conditions, further validating their effectiveness and adaptability in real-world scenarios. The integration of such advanced methodologies can significantly contribute to the sustainability and safety of road networks, ultimately benefiting both users and the environment.
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
