Software ensures safe bridges
Picture program recognizes and automatically marks material damageRead out
In Germany there are about 120, 000 bridges. To ensure that motorists, cyclists and pedestrians get across safely, they must be regularly checked for damage. Now, researchers have developed image processing software that automatically detects irregularities in the bridge material and can thus prevent accidents.
They span deep ravines, rivers, highways - bridges are essential for the transport network. However, their condition in Germany is frightening: in a 2007 ADAC study, one in ten of the fifty bridges tested failed the test; in total, four received the grade "poor" and one was even rated as "very poor". Changing weather and temperature influences, increasing traffic volumes and deicing salts cause stress on the material - damage such as hairline cracks, spalling concrete and rust through it quickly develops. If the bridge engineers do not recognize it in time, it will be dangerous for motorists, cyclists and pedestrians.
Image processing instead of adhesive tape
So far, the inspectors are investigating a bridge directly on site for externally visible damage. Cracks provide them with adhesive strips, which stretch as the crack grows larger. A new image processing program will make such control measures superfluous in the future. Researchers at the Fraunhofer Institute for Industrial Mathematics ITWM in Kaiserslautern have developed this software together with colleagues from Infracom Italy.
"The photos of a bridge are automatically examined by the software for certain properties and irregularities, such as strong color deviations, " explains Markus Rauhut, a scientist at the ITWM. "In contrast to humans, the tool overlooks no abnormalities - even the smallest damage is identified and marked." Display
Reference values stored in database
The challenge: no two bridges are alike. Shape, construction material and surface structure differ, the color is dependent on the material, the degree of moisture and dirt or fouling. The software must be able to handle these deviations. For this, the researchers have extracted metrics from photographs, such as the characteristic elongated shape of a hairline crack, the typical color deviations in wet areas or the structures of the material, which are different in a concrete bridge than a steel bridge. These deposited the scientists in a database.
When the researchers upload a photo to the program, the software compares the image properties of the new image with those of the saved image. If she discovers irregularities, she marks the respective area in the photo. The bridge checker can now decide how serious the damage is. Is there a need for action? The faster damage is detected and clearly classified, the cheaper and easier it is to repair. For about half a year, the engineers have been using the new software to successfully bridge in
(Fraunhofer-Gesellschaft, 04.11.2008 - DLO)