|
Measuring cracks on concrete bridge walls using video images
|
Farzad Zare zadeh   , Ali Hosseininaveh * , Mojtaba Mahmoodian |
|
|
|
Abstract: (7 Views) |
The identification and measurement of cracks in concrete structures, particularly bridges, are pivotal elements of health monitoring. Traditional assessment methods, while essential, present inconsistencies across individual experts and are laborious due to manual processes. Consequently, there is a pressing need for innovative approaches, such as image-based methods, that combine photogrammetry with deep learning techniques. This research introduces a novel image-based technique for detecting and measuring cracks in concrete bridges. The procedure commences with the capture of a video of the bridge, which is subsequently processed to extract key frames utilizing deep learning algorithms. Non-essential frames are discarded based on the criteria of image end lap. Following this, the projective centers of the target areas on the bridge are identified through the use of key frames. Subsequent to determining the projective centers, images pertinent to each crack are grouped together via clustering algorithms. This process refines the data set, enabling the production of orthophotos that depict the cracks. The dimensions of these cracks are then measured within the orthophotos, with reference objects employed to convert pixel measurements into physical units. Empirical evaluations demonstrate that this method significantly reduces the quantity of required images by approximately one-fifth while maintaining high accuracy. Consequently, the processing velocity is amplified by a factor of five. Additionally, comparison of the obtained results with actual values showed that the highest measurement error belongs to crack width of 2.1 mm and crack length of 11 mm. Therefore, the error in the presented method is approximately 3 to 4 percent. |
Article number: 3 |
| Keywords: Photogrammetry, deep learning, image processing, concrete bridge cracks, keyframes |
|
|
|
Type of Study: Research |
Subject:
Photo&RS
Received: 2024/02/6 | Accepted: 2025/12/18
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
