Sie können Operatoren mit Ihrer Suchanfrage kombinieren, um diese noch präziser einzugrenzen. Klicken Sie auf den Suchoperator, um eine Erklärung seiner Funktionsweise anzuzeigen.
Findet Dokumente, in denen beide Begriffe in beliebiger Reihenfolge innerhalb von maximal n Worten zueinander stehen. Empfehlung: Wählen Sie zwischen 15 und 30 als maximale Wortanzahl (z.B. NEAR(hybrid, antrieb, 20)).
Findet Dokumente, in denen der Begriff in Wortvarianten vorkommt, wobei diese VOR, HINTER oder VOR und HINTER dem Suchbegriff anschließen können (z.B., leichtbau*, *leichtbau, *leichtbau*).
Dieses Kapitel befasst sich mit der entscheidenden Notwendigkeit einer effizienten Überwachung der Risse auf den Gehwegen, um die Sicherheit und Langlebigkeit der Straße zu gewährleisten. Es untersucht automatisierte Methoden zur Erkennung und Charakterisierung von Rissen mithilfe von Bildverarbeitungstechniken wie Gradientenfiltern und neuronalen Netzwerken. Die Studie hebt die Entwicklung einer Modellplattform hervor, die es den Nutzern ermöglicht, Bilder von Straßenbelägen hochzuladen, was eine detaillierte Risscharakterisierung ermöglicht. Darüber hinaus wird die Integration von Georeferenzierungstools diskutiert, um die Funktionalität der Plattform zu verbessern. Das Kapitel vergleicht auch traditionelle manuelle Inspektionen mit automatisierten Lösungen und betont die Vorteile der letzteren. Experimentelle Ergebnisse zeigen die Effektivität der Kombination von konvolutionalen neuronalen Netzwerken (CNNs) mit Bildverarbeitungsfiltern zur präzisen Erkennung von Rissen. Die Entwicklung einer eigenen Website zur Bildersammlung und -charakterisierung unterstreicht den praktischen Nutzen dieser Forschung. Zukünftige Arbeiten zielen darauf ab, die Leistung dieses Ansatzes in realen Szenarien zu messen, um seinen praktischen Nutzen und seine Fähigkeiten zu bewerten.
KI-Generiert
Diese Zusammenfassung des Fachinhalts wurde mit Hilfe von KI generiert.
Abstract
Road pavements are vital for transportation infrastructure, yet they deteriorate over time due to traffic loads and environmental factors, resulting in cracks and damage. This paper introduces an innovative method for crack detection on road pavements using digital imagery. Our approach incorporates geo-localization, annotates, characterizes, and quantifies crack severity. This empowers experts to monitor crack progression, a critical element in pavement management. The methodology allows for seamless result comparison and augments existing techniques, aiding in condition assessment and conservation strategy determination. Timely detection of cracks enables proactive maintenance, preventing structural degradation, and ensuring user safety and comfort. Leveraging deep learning and open-source frameworks like TensorFlow and QGIS, our approach automates road pavement image analysis and crack identification, providing a cost-effective, accessible solution for crack detection. This research offers significant advantages in resource efficiency and accessibility, especially in areas without regular manual inspections or dedicated vehicles, thereby enhancing road pavement monitoring and maintenance.
1 Introduction
The global road network is essential for transportation, connecting communities and facilitating economic activities. However, maintaining road pavements in optimal condition is an ongoing challenge. Over time, roads are subjected to heavy traffic loads and environmental agents, leading to the development of cracks. These pavement distresses not only compromise the safety and comfort of road users but also accelerate the structural degradation of the pavement itself. Therefore, effective monitoring and timely maintenance are critical components of road life cycle management. In this work we intend to study how an automatic approach, with the help of image processing, can complement this monitoring, especially in the detection, characterization and level of degradation of cracks. Automatic methods will be analyzed, which enable the detection of cracks in road pavements. Techniques to be studied include image processing techniques, using gradient filters and neural networks. It is also intended to develop a model platform that allows users to provide images of road pavements, enabling the user to provide a detailed characterization of the cracks present in them, according to the available manuals. These tools must be used and validated by experts in the field. Finally, we intend to study a way to improve this platform by adding a tool that allows the georeferencing of images uploaded to the platform.
2 Related Work
Several methods have been proposed for crack detection in road pavements. Traditional approaches rely on manual inspections, which are time-consuming and often impractical for large road networks [1]. Automated solutions, such as computer vision-based methods and machine learning algorithms, have shown promise results in recent years.
Anzeige
2.1 Study of Pavements
Regarding their constitution, pavements are considered structures made up several horizontal layers of finite thickness, supported on a foundation of natural pavement Fig. 1-A. The main objective of these is to support the actions induced by vehicles, transmitting them to the foundation in an attenuated form, thus providing a safe and comfortable surface for vehicles to circulate [2].
Fig. 1.
(A) Schematic of the typical structure of road pavements and (B) different types of cracks. Both images produced using Libreoffice, an open source software.
There are several types of pavements, flexible pavement (the vast majority of pavements), rigid, and semi-rigid, each of these types, with different characteristics.
Degradation Mechanisms, cracks. Road pavements suffer, throughout their service, loss of load capacity, together with environmental causes (e.g. weathering and aging) which leads to the appearance of degradation. The main degradation mechanisms are the cracking of bituminous layers. There are several types of cracking, the most common is fatigue cracking. In addition to the cracks being characterized correctly, the aim is to provide more detailed information about them using the Road Pavement Degradation Catalog [9, 10], which can be done, as there is a section on the various types of degradation in flexible pavements and their degradation levels. We will focus on 3 levels of degradation of longitudinal, transverse cracks and alligator skin (Fig. 1-B).
2.2 Image Processing
Images of road pavements can be analyzed and improved using different filters. GIMP, GNU Image Manipulation Program, is an open source image manipulation program that can includes most major image processing filters. GIMP is ideal for our use case since it can be used both manually and fully automatically from a python program. For the present study GIMP was used to highlight the cracks that were presented in the images. It can often be seen that there is a crack, but its contour is not prominent. Edge Detectors, as the name suggests, are detection filters that look for edges between different colors, thus detecting the contours of objects present in the images. Below are some relevant filters from that are considered relevant for processing these images.
Anzeige
Difference of Gaussians. This filter detects edges, this performs two Gaussian blurs with different radius on the image. The radius can be configured, if the radius 1 is increased, the algorithm recognizes edges that appear thicker, if the radius 2 is decreased, the algorithm detects thinner edges.
Sobel. This type detects horizontal and vertical edges through the calculation of first-order derivatives, creating an intensity gradient [3]. The final result is an image with some level of transparency with black lines and some colors as can be seen in Fig. 2.
Dilate. The filter Dilate, enlarges and perfects the dark areas of the layer. For each pixel of the image, a pixel value (brightness) will be given equal to the lowest pixel value (the darkest) of 8 neighboring pixels in a matrix. Thus, a dark pixel is added if the pixels around it are also dark. This will cause that area to be dilated in all directions, creating a “big pixel”.
Erode Noise. This filter is part of the GIMP generic filters, it consists of enlarging and highlighting bright areas in an image.
Fig. 2.
Example of applying the Sobel filter in GIMP—Simpsons contributor, CC BY-SA 3.0, via Wikimedia Commons, 2025 ( Valve original (1) & Valve sobel (3)).
Deep Convolutional Neural Networks (CNNs) have achieved innovative results over the last decade in a variety of fields related to pattern recognition, image processing and speech recognition [6]. Convolutional Neural Networks (CNNs) automatically learn convolution filters, whereas the sobel filter mentioned earlier serves as an example of a manually designed convolutional filter.
VGG-16 is an architectural model of a CNN, which was developed to classify images. The “16” in VGG-16 refers to the number of layers the network contains. The paper [7] proposed a network based on 16 convolution layers and integrates a method to predict pixel segmentation and to improve the accuracy of crack detection, avoiding the data augmentation. For this study, images of cracks collected by UAVs and images of buildings in Sydney, Australia were used. There is also another study [8] that uses VGG-16 as a basis for developing the model for identifying cracks in concrete. This model was proposed by the ImageNet Large Scale Visual Recognition Challenge in 2014, and was trained on ImageNet Dataset, which consists of millions of images divided into thousands of categories. Cha et al. [4] developed a classic CNN to detect cracks in concrete surfaces and compared the results obtained with CNN to methods with edge detection tools, such as Canny and Sobel. The results of this study showed that the CNN had better performance in finding cracks in realistic situations [4]. Azimi, Mohsen et al. [5] state that, normally, two approaches are used to detect cracks in structures. One based on the image binarization section and the sequential image processing method.
4 Experimental Results
Summary of GIMP Filter Survey and Results. After carrying out a study on the filters using GIMP and which of these would be appropriate for detecting cracks in images of road pavements, different images such as Fig. 3 were analysed. Another approach was attempted. Instead of detecting just one crack in the image, it aimed to detect all the cracks in an original road pavement image. After identifying all the cracks in the image, several GIMP filters, including Erode and Dilate were applied multiple times. From this study, a crack was selected from the top of the image and an attempt was made to improve the detection of the crack by applying more filters. However results are still far from what is needed for our goal.
Results Using CNN and Image Processing. Hybrid strategies have been used to improve and increase the automatic detection of cracks in road pavements [1]. Like the approach proposed, a combination of the use of a neural network and image processing to detect cracks in road pavements Fig. 3.
Fig. 3.
Results obtained. The first is the crack identified on red, followed by the second imagem on the right, that was obtained by using the filters described above. On the third and fourth image is presented two isolated cracks. Finally the last imagem is the result obtained by using the VGG-16 and image processing. Images produced using GIMP and Python, both open source software. (Color figure online)
User Connection. In order to connect the user to this approach, a website was developed. The user is able to upload photos of road pavements taken with cameras or phones. The user can also make a brief description of the crack and evaluate the type based on the descriptions and imagens presented on the website, insert the coordinates where the photo was taken and insert the name and email. Figure 4 presents some prints of the developed website: https://falhaavista.web.app.
5 Conclusions and Discussion
The presented crack detection method offers significant advantages. It is highly accessible, requiring minimal resources for data collection, where standard cameras or drones can be employed, making it a cost-effective solution. Moreover, the automation of crack detection reduces the reliance on manual inspections, particularly valuable in regions where dedicated inspection campaigns are challenging.
This paper introduces an innovative approach to road pavement crack detection using digital imagery and deep learning techniques [1]. This method provides a cost-effective and accessible means of monitoring road pavements, enhancing their durability and road user safety. The development of a dedicated website for collecting images and aiding in their characterization further emphasizes the potential of this research. Future work will focus on measuring the performance of this approach in real-world crack pavment characterization to assess its practical utility and capabilities.
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Cha, Y.-J., Choi, W., Oral, Deep Learning-Based Crack Damage Detection Using Convolutional Neural Networks. In: Computer-Aided Civil and Infrastructure Engineering (2017). https://doi.org/10.3390/drones6010005
5.
Azimi, M., Eslamlou, A.D., Pekcan, G.: Data-Driven Structural Health Monitoring and Damage Detection through Deep Learning: State-of-the-Art Review, in Sensors (2020). https://doi.org/10.3390/s20102778
6.
Krizhevsky, A., Sutskever, I., Hinton, G.E.: ImageNet classification with deep convolutional neural networks. In: Neural Information Processing Systems (2012). https://doi.org/10.1145/3065386
7.
Munawar, H.S., et al.: Inspecting Buildings Using Drones and Computer Vision: A Machine Learning Approach to Detect Cracks and Damages, Advances in Neural Information Processing Systems (2022). https://doi.org/10.3390/drones6010005
8.
Silva, Wilson Ricardo Leal da and Lucena, Diogo Schwerz de, Concrete Cracks Detection Based on Deep Learning Image Classification, in Proceedings, 2018. https://doi.org/10.3390/ICEM18-05387
9.
CEPSA, Manual de Pavimentação (2006)
10.
EP Estradas de Portugal, Catálogo de Degradações dos Pavimentos Rodoviários, vol.2: Gestão da Conservação (2008)
