The project „Optical 3D Bridge Inspection”, which is part of DFG’s priority program “100+”, aims to capture surface geometry and damages of prestressed concrete bridges with high-resolution optical tools to assist the structural health monitoring process. In a multi-scale and multi-epoch approach, the building structure is recorded in total with UAV-based cameras in millimetre-resolution – extracting deformation and areas of interest where damages are visible – and in those hotspots, an even higher resolved image block and also a micrometre-resolution structured light scanner (SLS) capture is taken. Detected damages are compared between multiple epochs, monitoring their development. In an experiment, we demonstrated all measurements and their linking possibilities on a reinforced concrete plate under controlled load. The global image block showed that even the smallest cracks with a width of 0.05mm were visible in the images with a spatial resolution of 0.16mm per pixel. Also, the three-dimensional reconstruction based on the images was able to mirror the plate accurately in all epochs. However, it was shown that manually applied speckles reduced the noise drastically, compared to areas in which the surface only consisted of blank light concrete with only a few microfeatures. The underlying deformation was nevertheless accurately reconstructed in all areas. It is shown that the SLS can measure the width and detailed shape of a crack, which enabled us to track changes between multiple epochs of different load. With feature-based matching of photogrammetry and SLS results, we can combine the advantages of the fast global UAV-based image approach and the micrometer-resolution local SLS scans and use them to support each other. In this paper we report about those experiments in detail and analyze them. Based on that, we formulate a possible user story to apply both techniques to support structural health monitoring of bridges and decision-making in maintenance.

Container cranes are of key importance for maritime cargo transportation. The uninterrupted and all-day operation of these container cranes, which directly affects the efficiency of the port, necessitates the continuous inspection of these massive hoisting steel structures. Due to the large size of cranes, the current manual inspections performed by expert climbers are costly, risky, and time-consuming. This motivates further investigations on automated non-destructive approaches for the remote inspection of fatigue-prone parts of cranes. In this paper, we investigate the effectiveness of color space-based and deep learning-based approaches for separating the foreground crane parts from the whole image. Subsequently, three different ML-based algorithms (k-Nearest Neighbors, Random Forest, and Naive Bayes) are employed to detect the rust and repainting areas from detected foreground parts of the crane body. Qualitative and quantitative comparisons of the results of these approaches were conducted. While quantitative evaluation of pixel-based analysis reveals the superiority of the k-Nearest Neighbors algorithm in our experiments, the potential of Random Forest and Naive Bayes for region-based analysis of the defect is highlighted.

Container crane inspection is a very important task to maintain their uninterrupted operation. Nevertheless, this is a costly and time-consuming activity if performed manually. Recently, image-based detection of surface damages or changes using drones has gained increasing interest in industry; especially when objects of interest have a complex structure like container cranes. One main aim of this paper is a single-epoch image analysis which will also serve later for multi-epoch processing. It provides reliable information about current defects that may lead to big damages if not inspected by experts. Naïve Bayes classifier is employed to classify the images in different classes of which critical defects and especially rust is important. The preliminary results show that the precision on the target class reached about 99%. However, 87% percent recall in this class is not enough and it should be improved for this application.Having a large dataset requires an efficient data management system to provide users and decision makers with the information needed. In addition, in order to foster full automation, the aforementioned image analysis component should have a direct connection to the database and thus is able to query image and semantic information. We therefore introduce the second aim of our research, that is a concept for database design. Here, not only the raw data and the final results are integrated but also the intermediate results. At the same time, the database concept is connected to an integrated client interface that allows retrieving data of interest in a virtual globe.

The automotive industry is expanding its efforts to develop new techniques for increasing the level of intelligent driving and create new autonomous cars capable of driving with more intelligent capabilities. Thus, companies in this sector are turning to the development of autonomous cars and more specifically developing software along with more artificial intelligent algorithms. However, to be able to trust these systems, they must be developed very carefully, and use techniques that can increase the level of recognition that will consequently improve the level of safety. One of the most important components in this respect for road users is the correct interpretation of traffic sings. This paper presents a deep learning model based on convolutional neural networks and image processing that can be used to improve the recognition of traffic sings autonomously. The results are focused on difficult cases such as images with lighting problems, blurry traffic sings, hidden traffic sings, and small images. Hence, real cases are used in this study for identifying the existing problems and achieving good performance in traffic signal recognition. Finally, as a result, the configuration of the neural architecture based on three phases of convolutions proposed shows a validation accuracy of 99.3% during the data training. Another comparison carried out with the model ResNet-50 obtained an accuracy of 88.5%. Thus, for this type of application, a high validation accuracy is required as the results of our model demonstrated.