A5073930300- Belt Conveyor Systems Engineering
- Mechanical and Thermal Properties Analysis
- Industrial Vision Systems and Defect Detection
- Optical measurement and interference techniques
- Video Surveillance and Tracking Methods
- Advanced Neural Network Applications
- Image and Object Detection Techniques
- Advanced Vision and Imaging
- Industrial Engineering and Technologies
- Engineering and Environmental Studies
- 3D Surveying and Cultural Heritage
- Plant Pathogens and Fungal Diseases
- Human Pose and Action Recognition
- Anomaly Detection Techniques and Applications
- Advanced Image and Video Retrieval Techniques
- Surface Roughness and Optical Measurements
- Metallurgy and Material Forming
- Domain Adaptation and Few-Shot Learning
- Robotics and Sensor-Based Localization
- Advanced Measurement and Metrology Techniques
- Image Processing Techniques and Applications
- Agricultural Engineering and Mechanization
- Autonomous Vehicle Technology and Safety
- Engine and Fuel Emissions
- Gaze Tracking and Assistive Technology
Karlsruhe Institute of Technology
2015-2024
Fraunhofer Institute of Optronics, System Technologies and Image Exploitation
2015-2024
The Ohio State University
2024
Weatherford College
2022
Japan Science and Technology Agency
2019-2022
Naver (South Korea)
2022
Walter de Gruyter (Germany)
2020-2022
Engineering Systems (United States)
2022
Information System Technologies (United States)
2021
Microsoft (United States)
2019-2021
Multi-pedestrian trackers perform well when targets are clearly visible making the association task quite easy. However, heavy occlusions present, a mechanism to re-identify persons is needed. The common approach extract visual features from new detections and compare them with of previously found tracks. Since those can have substantial overlaps nearby – especially in crowded scenarios extracted insufficient for reliable re-identification. In contrast, we propose novel occlusion handling...
Vehicle detection in aerial images is a crucial image processing step for many applications like screening of large areas. In recent years, several deep learning based frameworks have been proposed object detection. However, these detectors were developed datasets that considerably differ from images. this paper, we systematically investigate the potential Fast R-CNN and Faster images, which achieve top performing results on common benchmark datasets. Therefore, applicability 8...
Optimisation of manufacturing process parameters requires resource-intensive search in a high-dimensional parameter space. In some cases, physics-based simulations can replace actual experiments. But they are computationally expensive to evaluate. Surrogate-based optimisation uses simplified model guide the for optimised combinations, where surrogate is iteratively improved with new observations. This work applies surrogate-based composite textile draping process. Numerical experiments...
Recent progress in the development of unmanned aerial vehicles (UAVs) causes serious safety issues for mass events and safety-sensitive locations like prisons or airports. To address these concerns, robust UAV detection systems are required. In this work, we propose an framework based on video images. Depending whether images recorded by static cameras moving cameras, initially detect regions that likely to contain object median background subtraction a deep learning proposal method,...
Multi-person tracking is often solved with a tracking-by-detection approach that matches all tracks and detections simultaneously based on distance matrix. In crowded scenes, ambiguous situations similar track-detection distances occur, which leads to wrong assignments. To mitigate this problem, we propose new association method separately treats such difficult by modelling assignments the differences in Depending numbers of detections, for assignment task determined ambiguous, different...
Point cloud sampling is a less explored research topic for this data representation. The most commonly used methods are still classical random and farthest point sampling. With the development of neural networks, various have been proposed to sample clouds in task-based learning manner. However, these mostly generative-based, rather than selecting points directly using mathematical statistics. Inspired by Canny edge detection algorithm images with help attention mechanism, paper proposes...
Deflectometry as a technique to assess reflective surfaces has now existed for some 40 years. Its different aspects and variations have been studied in multiple theses research articles; reviews are available certain subtopics. Still field of active development with many unsolved problems, deflectometry encompasses large variety application domains, hardware setup types, processing workflows purposes, spans range from qualitative defect inspection vehicles precision measurements microscopic...
Recognizing soft-biometric pedestrian attributes is essential in video surveillance and fashion retrieval. Recent works show promising results on single datasets. Nevertheless, the generalization ability of these methods under different attribute distributions, viewpoints, varying illumination, low resolutions remains rarely understood due to strong biases current To close this gap support a systematic investigation, we present UPAR, Unified Person Attribute Recognition Dataset. It based...
Although human action anticipation is a task which inherently multi-modal, state-of-the-art methods on well known datasets leverage this data by applying ensemble and averaging scores of uni-modal networks. In work we introduce transformer based modality fusion techniques, unify multi-modal at an early stage. Our Anticipative Feature Fusion Transformer (AFFT) proves to be superior popular score approaches presents results outperforming previous EpicKitchens-100 EGTEA Gaze+. model easily...
To model increasingly adaptive production systems, skills are used to describe generic capabilities of the system components. In this paper, authors extend well-known division entities into product, process, and resource (PPR) with a skill definition. There two main advantages for approach: First, using PPR definition allows easy integration existing models tools. Second, there is natural tendency define very capture all possible use cases. But at some point, have be translated precise...
Abstract Shortening product lifecycles and small lot sizes require manufacturing systems to adapt increasingly fast. Many existing machine tools, handling logistics provide a generic functionality that is not bound specific product. But this flexibility reconfigurability on the level of individual resources lost in automated are limited production fixed set variants. We propose unified abstraction for skills provided by available product-specific requirements. From these high-level...
Wide Area Motion Imagery (WAMI) enables the surveillance of tens square kilometers with one airborne sensor Each image can contain thousands moving objects. Applications such as driver behavior analysis or traffic monitoring require precise multiple object tracking that is dependent on initial detections. However, low resolution, dense traffic, and imprecise alignment lead to split, merged, missing No systematic evaluation detection exists so far although many approaches have been presented...
In many visual surveillance applications the task of person detection and localization can be solved easier by using thermal long-wave infrared (LWIR) cameras which are less affected changing illumination or background texture than visual-optical cameras. Especially in outdoor scenes where usually only few hot spots appear imagery, humans detected more reliably due to their prominent signature. We propose a two-stage recognition approach for LWIR images: (1) application Maximally Stable...
Multi-camera tracking of vehicles on a city-scale level is crucial task for efficient traffic monitoring. Most the errors made by such multi-target multi-camera systems arise due to failures or misleading visual information detection boxes under occlusion. Therefore, we propose an occlusion-aware approach that leverages temporal from tracks improve single-camera performance occlusion handling strategy and additional modules filter false detections. For tracking, discard obstacle-occluded...
Sensor-based sorting describes a family of systems that enable the removal individual objects from material stream. The technology is widely used in various industries such as agriculture, food, mining, and recycling. Examples tasks include fungus-infested grains, enrichment copper content mining or plastic waste according to type plastic. Sorting decisions are made based on information acquired by one more sensors. A particular strength flexibility decisions, which achieved using sensors...
Automated cooperative collision avoidance of multiple vehicles is a promising approach to increase road safety in the future. This requires real-time motion planner which computes maneuvers cognitive vehicles. As planning task high computational complexity, computing times have be traded off against solution quality. contribution compares several algorithms with respect these criteria. The considered are tree search algorithm relying on precomputed lower bounds, elastic band method,...
Imaging sensors capturing the surroundings of an autonomous vehicle are vital for its understanding environment. While thermal infrared cameras promise improved bad weather and nighttime robustness compared with standard RGB cameras, detecting objects, such as persons, in imagery is a tough problem because image resolution quality typically far lower, especially low-cost sensors. Currently, deep learning based object detection frameworks offer impressive performance on high-quality images....
Small drones are a rising threat due to their possible misuse for illegal activities, in particular smuggling and terrorism. The project SafeShore, funded by the European Commission under Horizon 2020 program, has launched "drone-vs-bird detection challenge" address one of many technical issues arising this context. goal is detect drone appearing at some point video where birds may be also present: algorithm should raise an alarm provide position estimate only when present, while not issuing...
Human pose estimation has recently made significant progress with the adoption of deep convolutional neural networks and many applications have attracted tremendous interest in recent years. However, these require for human crowds, which still is a rarely addressed problem. For this purpose work explores methods to optimize focusing on challenges introduced larger scale crowds like people close proximity each other, mutual occlusions, partial visibility due environment. In order address...
The rapidly increasing number of surveillance cameras offers a variety opportunities for intelligent video analytics to improve public safety. Among many others, the automatic recognition suspicious and violent behavior poses key task. To preserve personal privacy, prevent ethnic bias, reduce complexity, most approaches first extract pose or skeleton persons subsequently perform activity recognition. However, current literature mainly focuses on research datasets does not consider real-world...
Anticipating future actions is a critical yet challenging task essential for video understanding. Effective utilization of observed information from the past fundamental to making accurate predictions. While prior studies in action anticipation have predominantly centered on architectural enhancements optimize use historical data, this work, we propose new framework that formulates as denoising diffusion process noisy states actions. In framework, are initially drawn standard Gaussian noise...