Csaba Beleznai

• PhD
• Senior Researcher at Austrian Institute of Technology

Efficient material logistics play a critical role in controlling costs and schedules in the construction industry. However, manual material handling remains prone to inefficiencies, delays, and safety risks. Autonomous forklifts offer a promising solution to streamline on-site logistics, reducing reliance on human operators and mitigating labor sho...

TalkWithMachines aims to enhance human-robot interaction by contributing to interpretable industrial robotic systems, especially for safety-critical applications. The presented paper investigates recent advancements in Large Language Models (LLMs) and Vision Language Models (VLMs), in combination with robotic perception and control. This integratio...

A considerable amount of research is concerned with the challenging task of estimating three-dimensional (3D) pose and size for multi-object indoor scene configurations. Many existing models rely on a priori known object models, such as 3D CAD models and are therefore limited to a predefined set of object categories. This closed-set constraint limi...

The estimation of depth cues from a single image has recently emerged as an appealing alternative to depth estimation from stereo image pairs. The easy availability of these dense depth cues naturally triggers research questions, how depth images can be used to infer geometric object and view attributes. Furthermore, the question arises how the qua...

Automated monitoring and analysis of passenger movement in safety-critical parts of transport infrastructures represent a relevant visual surveillance task. Recent breakthroughs in visual representation learning and spatial sensing opened up new possibilities for detecting and tracking humans and objects within a 3D spatial context. This paper prop...

Automated monitoring and analysis of passenger movement in safety-critical parts of transport infrastructures represent a relevant visual surveillance task. Recent breakthroughs in visual representation learning and spatial sensing opened up new possibilities for detecting and tracking humans and objects within a 3D spatial context. This paper prop...

Human detection in crowded situations represents a challenging task in many practically relevant scenarios. In this paper we propose a passive stereo depth based human detection scheme employing a hierarchically-structured tree of learned shape templates for delineating clusters corresponding to humans. In order to enhance the specificity of the de...

This chapter provides an overview of acquisition methods, specific challenges, and solutions for image-based metal surface inspection. We discuss illumination and recording methods for a wide range of tasks such as inspection of rail surfaces, weld quality assessment, inspection of holograms on metallic foils, and reconstruction of coins. Algorithm...

Recently, Mahalanobis metric learning has gained a considerable interest for single-shot person re-identification. The main idea is to build on an existing image representation and to learn a metric that reflects the visual camera-to-camera transitions, allowing for a more powerful classification. The goal of this chapter is twofold. We first revie...

Real-Time Multimedia Content Analysis opens up exciting possibilities for accessing opinion-oriented arguments about regulations and dynamic policy changes. In this chapter, the authors present common methodologies and core technologies to analyse multimedia content from a practitioner's viewpoint, highlighting their primary impact, best practices,...

Reliable and timely detection of abandoned items in public places still represents an unsolved problem for automated visual surveillance. Typical surveilled scenarios are associated with high visual ambiguity such as shadows, occlusions, illumination changes and substantial clutter consisting of a mixture of dynamic and stationary objects. Motivate...

We present a GPU-accelerated, real-time and practical, pedestrian detection system, which efficiently computes pedestrian-specific shape and motion cues and combines them in a probabilistic manner to infer the location and occlusion status of pedestrians viewed by a stationary camera. The articulated pedestrian shape is approximated by a mean conto...

One central task in many visual surveillance scenarios is person re-identification, i.e., recognizing an individual person across a network of spatially disjoint cameras. Most successful recognition approaches are either based on direct modeling of the human appearance or on machine learning. In this work, we aim at taking advantage of both directi...

This paper summarises work undertaken on the VITUS project. The main aim of VITUS project is to build and implement a prototype for an automatic video image analysis system in order to increase safety in tunnel roads. A feasibility study about video image analysis in tunnels was carried out, and the implementation of the prototype and evaluation of...

Discovering frequent and rare spatio-temporal patterns in large amounts of streaming visual data is of great practical interest since it allows for automated applications of activity and surveillance analysis. In this paper we present a computationally efficient and memory preserving clustering scheme which uses streaming input from a stationary-mo...

Real-time pedestrian detection in crowded scenarios still represents a major scientific challenge. Dynamic occlusions between humans and the presence of dense gradient structure (clutter) typically render such scenarios complex for automated visual analysis. In this demo we present an algorithmic framework which efficiently computes pedestrian-spec...

Poly (tetrafluorethylene) and polyimide samples were irradiated by a pulsed laser source at 308 nm and the resulting surface morphology was investigated. The photoablated surfaces show a strong dependence on the optical and structural parameters of the polymers. The roughness of the fractal surfaces has been characterized by means of calculating th...

form only given. Anthropomatics addresses the symbiosis between humans and machines, focusing on a deeper understanding of the cooperation, interaction and coexistence between humans and machines stimulating and strengthen advanced and deep research in response to the challenges of increasingly smart environments and multimodal access to various co...

Achieving accurate pedestrian detection for practically relevant scenarios in real-time is an important problem for many applications, while representing a major scientific challenge at the same time. In this paper we present an algorithmic framework which efficiently computes pedestrian-specific shape and motion cues and combines them in a probabi...

Person re-identification, i.e., recognizing a single person across spatially disjoint cameras, is an important task in visual surveillance. Existing approaches either try to find a suitable description of the appearance or learn a discriminative model. Since these different representational strategies capture a large extent of complementary informa...

Surveillance archives encompass vast amount of data. Given the amount of data the need for search and data exploration arises naturally. Various authorities such as infrastructure operators and law enforcement agencies are confronted with search needs based on a visual description (size, color, clothing, number plates, facial biometry, etc.) and/or...

This paper presents a data-oriented tracking framework which aims to recover the spatio-temporal trajectories for an unknown number of interacting objects appearing and disappearing at arbitrary times. Data association is performed at three-levels of a hierarchy: (i) first, trajectory segments and an associated quality measure are generated by a lo...

In this paper a novel non-rigid matching technique is presented to reliably detect near-regular object configurations in images in the presence of substantial clutter. The objects in near-regular configurations span a grid with not necessarily congruent grid cells. In addition, the output of object detectors is typically associated with ambiguities...

Object tracking is an extensively researched subject within the field of computer vision, in reference to the large number of published contributions in major computer vision conference proceedings and journals every year. This chapter presents object tracking achievements over the past years as well as research trends. It attempts to give a struct...

Motion is a strong cue for the pedestrian detection task. Several motion detection approaches exist which segment moving foreground regions quite reliably, nevertheless, correct estimation of a class label for the segmented objects still represents a challenge. Certain object classes such as pedestrian groups and vehicles are difficult to discrimin...

The complexity of human detection increases significantly with a growing density of humans populating a scene. This paper presents a Bayesian detection framework using shape and motion cues to obtain a maximum a posteriori (MAP) solution for human configurations consisting of many, possibly occluded pedestrians viewed by a stationary camera. The pa...

Tracking of spatially extended targets with variable shape, pose and appearance is a highly challenging task. In this work we propose a novel tracking approach using an incrementally generated part-based description to obtain a specific representation of target structure. The hierarchical part-based representation is learned in a generative manner...

This paper presents an approach for video object segmentation. The main idea of our approach is to generate a planar, triangulated, and labeled graph that describes the scene, foreground objects and background. With the help of the Kanade-Lucas-Tomasi Tracker, corner points are tracked within a video sequence. Then the movement of the points adapti...

The tracking of storm centres in radar data is of particular importance for short term weather prediction and specifically thunderstorm prediction. This paper presents a method to track storm centres in terrestrial radar images. Mean shift segmentation is used to outline storm centres and mean shift tracking to locate the storm in the consecutive i...

This paper presents a system for road sign detection based on edge orientation histograms. Edge orientation histograms are reliable, scale and contrast invariant features that can be extracted efficiently using integral images. A learning method is introduced that selects features based on the implicit transmission function of the designer's templa...

Visual surveillance and activity analysis is an active research field of computer vision. As a result, there are several different algo- rithms produced for this purpose. To obtain more robust systems it is desirable to integrate the different algorithms. To help achieve this goal, we propose a flexible, distributed software col- laboration framewo...

Change detection by background subtraction is a common approach to detect moving foreground. The resulting difference image is usually thresholded to obtain objects based on pixel connectedness and resulting blob objects are subsequently tracked. This paper proposes a detection approach not requiring the binarization of the difference image. Local...

Tracking multiple interacting objects represents a chal- lenging area in computer vision. The tracking problem in general can be formulated as the task of recovering the spatio-temporal trajectories for an unknown number of ob- jects appearing and disappearing at arbitrary times. Ob- servations are noisy, their origin is unknown, generated by true...

Change detection by background subtraction is a common approach to detect moving foreground. The resulting difference image is usually thresholded to obtain objects based on pixel connectedness and resulting blob objects are subsequently tracked. This paper proposes a detection approach not requiring the binarization of the difference image. Local...

This paper describes a postprocessing algorithm for improving the performance of a detec-tion system, i.e. to increase the detection rate and simultaneously decrease the number of false positives. In order to increase the detection rate we propose to use an approach called wobble, i.e. applying small random affine transformations to the image and r...

Tracking multiple targets-such as humans in a busy scene-is a non-trivial task due to the frequent occlu-sions occurring between the target objects. This work describes a novel way to detect human candidates directly from the non-thresholded difference image obtained by background differencing and to track detected candidates by a fast variant of t...

Detecting individual humans within groups becomes a non-trivial task when performing automatic visual surveillance in crowded scenes. This paper proposes a novel way to detect individual humans directly from the difference image using a fast variant of the mean shift mode seeking procedure and verifying the hypothesized configuration by a model-bas...

This work introduces a methodology for evaluating the operational range of a video surveillance system in terms of robustness and reliability. We propose the generation of semi and full-synthetic video sequences under controlled variation of selected parameters. This data provides the necessary ground truth information for evaluating the motion det...

In this paper we describe a surveillance system that is not only able to detect blobs and track them but also determines if a blob is a person. The given blob is segmented into sub-regions. A person model is fit to these regions such that a likelihood measure is maximized. The likelihood measure depends on the number of identified body parts, their...

Fingerprint recognition and verification are often based on local fingerprint features, usually ridge endings or terminations, also called minutiae. By exploiting the structural uniqueness of the image region around a minutia, the fingerprint recognition performance can be significantly enhanced. However, for most fingerprint images the number of m...

Picosecond pulsed laser irradiation of several metallic photocathodes (W, Ta, Al and Au) was carried out along with work function measurements during the irradiation process, work function measurements were performed before irradiation and after different irradiation periods. Metals used in this study behaved differently according to their affinity...

By using laser-induced photoemission and complementary in situ monitoring of photoelectric response spectra, damage threshold values have been determined for three wavelengths (213, 266 and 532 nm) of a pulsed picosecond Nd-YAG laser. The photoelectric signal is extremely sensitive on the surface state of the diamond-like carbon (DLC) film, therefo...

In this work, we correlate the structural properties of DLC films with their photoelectric properties. DLC layers are deposited on silicon wafers by RF plasma-assisted chemical vapor deposition (RF PACVD). The films are irradiated with picosecond Nd3+-YAG laser pulses at different wavelengths (213 and 266 nm) and energies (few μJ to mJ). Photoelect...

Laser-induced surface damage introduces a dramatic change in the photoelectric properties of metallic and semiconductor substrates. Hence, by varying the applied laser intensity, ultra-short- pulsed laser-induced photoemission (in the mono- or multiphotonic regime according to the applied wavelength) can be used to monitor in situ surface structura...

K+-implanted W samples with various implantation depths were investigated. Generally, implantation of alkali ions gives rise to two competitive effects: it lowers the surface work function, however it enhances surface oxidation too which in turn leads to a slight work function increase. In opposite to alkali overlayers, implanted species confined w...

Photocurrent measurements yielded new data, which were used to determine the surface work function of a native oxide-covered tungsten photocathode. The photoemission was generated by a continuous UV source. The surface work function has been measured at various stages of a pulsed UV laser-induced oxide removal in order to characterize the process....

In-situ monitoring of the vacuum chamber pressure during laser-induced oxide desorption on W surfaces provides a view on the desorption process. Sudden pressure variations suggest a change in the rate of desorption. During the desorption experiment, due to the incident pulsed UV laser light, photoelectrons are ejected from the sample surface. Evalu...

Results of UV (308 nm) laser pulse induced dry etching with subsequent Pd deposition from a PdCl2 solution (acid base with pH=1) on polyimide surface are reported. The surface roughness has been determined before and after illumination. The fractal-based examination techniques based on the area–perimeter and the structure function methods. It could...

In the past years image analysis has gained great relevance in several fields of scientific application. With these systems an affordable price might be limiting factor for small-scale research projects or educational purposes. The system presented here provides an excellent performance- price ratio and is thus an excellent choice for low-budget ap...

Experimental results of laser assisted chemical vapor deposition of nickel from Ni(CO)4 and theoretical treatment of deposition process are presented. The nickel deposition has ben realized by scanning of Ar+ laser beam (100 - 400 mW, (lambda) equals 515 nm and 488 nm) on Si surfaces in atmosphere of Ni(CO)4 with 0.2 - 2.0 mbars with scanning speed...

Ultrafast changes in the crystal structure of GaAs induced by intense femtosecond laser pulses are detected and investigated. Atomic force microscopy and Raman microprobe analysis of the laser-treated area show centrosymmetric (disordered) features which are different from the original zinc-blend structure of the GaAs lattice. The frozen-in structu...

A native oxide film on a metallic surface corresponds generally to a multilayer. We describe a method using a number of picosecond laser pulses at 213 nm to achieve partial removal of the oxide overlayer on a polycrystalline W substrate under UHV conditions. The process of oxide removal was monitored by measuring the charge of emitted photoelectron...

Polycrystalline metallic photocathodes with a covering native oxide layer were exposed to several wavelengths (213, 266 and 355 nm) of a picosecond Nd3+-YAG laser. Laser illumination at each wavelength was used to induce oxide film removal and a measurable photoelectric response simultaneously. The change of the photoelectric signal was monitored t...

The fifth harmonic (213 nm) of a picosecond Nd3+ :YAG laser was used to remove small fractions of the covering oxide overlayer from polycrystalline metallic photocathodes. The evolution of the photoemitted charge (high-density electron pulses) was followed along with the surface reflectivity change, and they were interpreted in terms of the work fu...

Experiments performed in the field of IR laser-induced oxidation of metals and their theoretical explanations are reported. Temperature variations in time of irradiated metallic samples have been recorded with a PC based system capable of evaluating the absorptivity variation function (A(T)) from recorded data. Measurements have been carried out on...

The optical properties of fullerenes embedded in other media is explored, assuming that a single fullerene behaves as a microscopic crystallite. The justification for this assumption is discussed, and various experimental results are compared to test the limits of this approximation. The effective medium theory is used to calculate the optical resp...

The task of reliable detection and tracking of multiple objects becomes highly complex for crowded scenarios. Data association is difficult to perform reliably in the presence of missing observations due to occlusions. We propose a novel real-time approach to segment and track multiple overlapping humans. The optimal segmentation solution is given...

Acquiring a highly specific target representation is a major challenge in the task of visual object tracking. High specificity substantially lowers the inherent ambiguity of the data association task and leads to improved tracking accuracy in presence of clutter. In this paper we propose a method generating a specific representation of the image st...

Change detection by background subtraction is a common approach to detect moving foreground. The resulting difference image is usually thresholded to obtain objects based on pixel connectedness and resulting blob objects are subsequently tracked. This paper proposes a detection approach not requiring the binarization of the difference image. Local...

This paper summarises work in progress of VITUS project. The main aim of VITUS project is to build and implement a prototype for an automatic video image analysis system in order to increase safety in tunnel roads. To achieve their objectives, VITUS is divided into two subprojects called VITUS-1, and VITUS-2 respectively. The former is a feasibilit...