Three-dimensional reconstruction from scanning electron microscope (SEM) micrographs
SEM micrographs allow the visualization of objects that can't be sufficiently magnified by conventional light microscopes. Nowadays modern devices achieve resolutions in the sub-nanometer range by using cold fieldemission cathodes and provide an excellent depth of focus and image quality. Nevertheless a crucial disadvantage is the fact that SEM micrographs don't contain information about the three-dimensional sample texture. Especially in investigations of particles and their morphology, depth maps are essential to generate a three-dimensional model of the specimen.
In the context of this project, the Institute of Measurement and Automatic Control is developing a method that allows the euclidean reconstruction of different-sized particles from SEM stereo micrographs. The method is based on the principles of epipolar geometry. Using a eucentrically tiltable specimen stage, a set of stereo images can be created. After the rectification of the image pairs, correspondences are determined using local and global matching algorithms. In the next step a disparity map can be computed that allows the triangulation of a point cloud and hence the computation of a projective reconstruction. Due to the uncalibrated scenery, the result isn't unambiguous and therefore it differs from an euclidean reconstruction by an unknow projective transformation. To determine this transformation, several approaches exist and thus a quantitative three-dimensional measurement of the examined specimen can be realized.
Contact person: Dipl.-Ing. Stefan Töberg
HyMon - Hand hygiene monitoring in hospitals
Several studies since the mid-19th century indicate the impact of hand hygiene compliance on the decline of infections in the health sector. However, despite this knowledge and worldwide-accepted guidelines on hand hygiene, infection rates in today’s hospitals are still too high. E.g., the number of deaths related to infections with multi-resistant pathogens is 15,000 persons per year in Germany. Studies indicate an average compliance rate of hand hygiene guidelines of 41.2% – 55.2%. Required hand disinfection steps are either omitted or performed too shortly. E.g., the actual time spent during the hand disinfection process is often below 15 seconds, although the norm EN 1500 proposes 30 seconds.
The Institute of Measurement and Automatic Control is developing a system, which evaluates the quantity and quality of hand disinfection processes using an optical sensor and computer vision. The process of hand disinfection can be divided into sequences of gestures. The system is learning to distinguish those gestures during a training phase, which is commonly known as machine learning. Attributes such as gesture duration, kind, quantity and succession are recognized and evaluated. Finally, the system aims to support the user in real time at the hand disinfection process and informs about potential untreated areas.
Contact person: M. Sc. Armin Dietz
Early recognition system for playing-related musculoskeletal disorders of professional musicians
Disorders, caused by unphysiological postures of the human body, are highly-relevant for our current and future society. Endangered are specialist working at a construction site or office, who lift heavy materials or sit on chairs respectively.
For professional musicians preventing disorders caused by unphysiological postures or movements are an extra-ordinary challenge. In the context of this study, it is focused on instrumentalists. For playing composition, like the 6th Paganini etudes with 1800 notes a minute – approximately 30 notes per second, musicians are expected to cross the limits of human capability. Such disorder can cause handicaps in musician‘s professional life or in worst case irreversible damage and the end of their musical career.
Therefore, the Institute of Measurement and Automatic Control is developing an early warning system for disorders induced by posture deficiencies with the help of optical and acoustical sensors. Based on Machine Learning, Object Detection, Pose Estimation and Motion Capturing, the instrumentalist, instrument and relative position and movements between them are recognized. If an unphysiological posture is detected, the musician will receive a direct feedback and is able to improve it instantly.
Contact person: M. Sc. Benjamin Fehlandt
The intelligent operation room
Today's operating room (OR) is an environment for high tech equipment, used for therapeutic and diagnostic applications. The equipment is applied to treat the patient with new and better surgical techniques. This causes an increasing variety and complexity of surgical procedures which poses a further requirement on the staff. This is taken into account through intelligent functions, which should decrease the workload. Processes within the OR are new designed and optimized, to ensure the best patient treatment.
Within the shown laboratory, approaches in human machine interaction, patient positioning and collision control are evaluated. 3D sensors are used to observe the scene and algorithms of digital image processing are applied. Based on the results, the OR should achieve information about the OR state and situation in order to interact with the staff. Besides assisting functions, active manipulation of OR equipment through the interface is possible.
Our technical focus is in the development and implementations of sensor array systems. Because the OR- environment is very complex, due to many people and objects within a small room, the scene is observed from various viewpoints. Through this, the complete segmentation of objects becomes possible. This causes a lot of challenges in the fields of real time applicability, data fusion capability and consideration of complex picture failures, like reflection, shadowing and interfacial effects.
Contact person: M. Sc. Armin Dietz
Development of a newly image derotator
The analysis of the dynamic behaviour of rotary objects under operating conditions is a great challenge due to the overlayed objects rotation.
At the Institute of Measurement and Automatic Control (IMR) a novel image derotator with a specially designed mirror arrangement is developed, that can simplify the analysis of fast rotating objects. By means of a laser doppler vibrometer or image processing the analysis of vibration or deformation can be accomplished.
Contact person: M. Sc. Bettina Altmann