Conference

Computed tomography allows a look inside. It has become indispensable in medicine and is used for routine examinations of humans. But it can also be used to X-ray composite materials such as reinforced concrete and examine them non-destructively. For example, what happens inside a concrete component when it is subjected to mechanical stress? Where do cracks form? What is their nature? How do cracks behave under increasing load? These are the questions the 85th Heidelberg Image Processing Forum is asking.

This year, our institute is the host. Experts will present advances in image processing. In addition, there will be an exhibition and a tour at the University of Kaiserslautern-Landau. 

One approach for anomaly detection is to use Machine Learning to automatically detect cracks. This model can identify fine cracks even on heterogeneous backgrounds, even if they are only one pixel thick. In 3D images, cracks are often only slightly delineated from their complex surroundings, making reliable detection and segmentation difficult.

In our project DAnoBi (Detection of Anomalies in Image Data), which is founded by the Federal Ministry of Education and Research, we are working closely with partners to develop specialized methods. These should be able to find and precisely segment cracks even in very large images of up to 400GB. In this way, we aim to significantly improve the quality and efficiency of concrete inspection and maintenance.