Wyniki wyszukiwania - BazTech

1

Skeleton-based human action/interaction classification in sparse image sequences

Piwowarski Paweł, Kasprzak Włodzimierz

Journal of Automation Mobile Robotics and Intelligent Systems

|

2023

|

Vol. 17, No. 3

1--14

EN

Research results on human activity classification in video are described, based on initial human skeleton estimation in selected video frames. Simple, homogeneous activities, limited to single person actions and two-person interactions, are considered. The initial skeleton data is estimated in selected video frames by software tools, like “OpenPose” or “HRNet”. Main contributions of presented work are the steps of “skeleton tracking and correcting” and “relational feature extraction”. It is shown that this feature engineering step significantly increases the classification accuracy compared to the case of raw skeleton data processing. Regarding the final neural network encoder‐classifier, two different architectures are designed and evaluated. The first solution is a lightweight multilayer perceptron (MLP) network, implementing the idea of a “mixture of pose experts”. Several pose classifiers (experts) are trained on different time periods (snapshots) of visual actions/interactions, while the final classification is a time‐related pooling of weighted expert classifications. All pose experts share a common deep encoding network. The second (middle weight) solution is based on a “long short‐term memory” (LSTM) network. Both solutions are trained and tested on the well‐known NTU RGB+D dataset, although only 2D data are used. Our results show comparable performance with some of the best reported LSTM-, Graph Convolutional Network- (GCN), and Convolutional Neural Network-based classifiers for this dataset. We conclude that, by reducing the noise of skeleton data, highly successful lightweight- and midweight-models for the recognition of brief activities in image sequences can be achieved.

2

Kinovea analysis of high-speed video recording to determine kinematics of double pendulum in the long time scale

Wojewoda Jerzy

Bulletin of the Polish Academy of Sciences. Technical Sciences

|

2023

|

Vol. 71, nr 6

art. no. e147918

EN

Measurement of position and velocity of rotating objects relies on installation of additional devices, which can significantly change their dynamic properties. Non-contact methods appear not to have the above-mentioned drawback. To determine the angular kinematics, a video measurement technique stands as a non-contact alternative. The rotational motion can be recorded with a high-speed camera and then analyzed with free and open-source tracking software which allows one to detect and digitize positions of chosen markers and then to calculate angular positions of selected elements. Differentiation process determines rotary speed values. Analysis of long-term dynamical behavior by recording data visualized as position maps which possess half of the information usually stored in the well-known Poincaré maps is proposed.

3

Analiza ruchu pieszych na podstawie danych z automatycznego systemu rozpoznawania pojazdów i pieszych. Cz. 1, Opis badań i analiza obrazów wideo

Szulecki Artur, Szagała Piotr

Drogownictwo

|

2021

|

nr 11-12

295--308

PL

W artykule opisano metody badawcze oraz przedstawiono technologie i narzędzia, za pomocą których starano się zarówno zebrać, przechowywać, przetwarzać, jak i analizować dane dotyczące ruchu na przejściach dla pieszych. Przedstawiono podstawowe charakterystyki ruchu pieszego zebrane na podstawie obrazu z systemu automatycznego rozpoznawania pieszych i pojazdów. Wykonana analiza udowodniła istnienie różnic pomiędzy zachowaniem pieszych przechodzących w grupie i pojedynczo. Jako grupę traktować należy pieszych, którzy przechodzą przez jezdnię w tym samym kierunku przynajmniej w liczbie dwóch osób.

EN

This paper describes the research methods and presents the technologies and tools by which it sought to both collect, store, process and analyze the data concerning traffic at pedestrian crossings. The basic characteristics of pedestrian traffic collected from the automatic pedestrian and vehicle recognition system are presented. The analysis proved the existence of the differences between the behavior of the pedestrians crossing in a group and those crossing separately. A group should be considered to be pedestrians who cross the road in the same direction in the number of at least two people.

4

Reconstruction of traffic situations from digital video-recording using method of volumetric kinetic mapping

Kolla Eduard, Ondruš Ján, Vertal' Peter

Archiwum Motoryzacji

|

2019

|

Vol. 84, nr 2

147--170

EN

In the past the traffic accident reconstruction was based in principle only on indirect methods that use accident marks and witness reports. These data were then used within backward reconstruction of event for determination of motion status of accident participants and expression of desirable quantities as are initial velocities, impact velocities, distances travelled or temporal conditions. These methods and their accuracy are dependent on the width of intervals of input quantities or on limited possibilities for motion synchronization between numerable participants of road traffic accidents. Existence of footage from CCTV cameras that captures traffic situations presents very valuable source of information about these accidents. The goal of the article is proposal of a volumetric kinetic mapping (VKM) method of accident reconstruction from CCTV footage. The method is based on synthesis of videoediting, videoanalysis and kinetic simulation using dedicated software for accident reconstruction. The method was furthermore validated for speed-time and distance-time variables by means of experimental test runs and by subsequent application of the method in the reconstruction of these tests using PC-Crash simulation software and two videoediting software packages. Results of the reconstructions of validation runs using VKM method were then verified by comparing them to the measured data from Corrsys Datron Microstar measuring system.

5

Evaluation of face detection algorithms for the bank client identity verification

Szczodrak M., Czyżewski A.

Foundations of Computing and Decision Sciences

|

2017

|

Vol. 42, No. 2

137--148

EN

Results of investigation of face detection algorithms efficiency in the banking client visual verification system are presented. The video recordings were made in real conditions met in three bank operating outlets employing a miniature industrial USB camera. The aim of the experiments was to check the practical usability of the face detection method in the biometric bank client verification system. The main assumption was to provide a simplified as much as possible user interaction with the application. Applied algorithms for face detection are described and achieved results of face detection in the real bank environment conditions are presented. Practical limitations of the application based on encountered problems are discussed.

6

Determining the shift of a bronchoscope catheter from the analysis of a video sequence of a bronchoscope video camera

Michalski D., Tabor Z., Zieliński B.

Biocybernetics and Biomedical Engineering

|

2017

|

Vol. 37, no. 4

630--636

EN

In this study we have proposed an algorithm for automated monitoring of the movements of a catheter used in peripheral bronchoscopy examination. We have shown that the shift of the catheter can be controlled in an automated way with quite a good accuracy by the means of analysis of video sequence recorded by a video camera of a bronchoscope. For a catheter moving between successive frames by no more than 1/3 of the distance between successive markers associated with a catheter the accuracy of a catheter shift measurement was equal to 1% and for a catheter moving between successive frames by no more than 1/2 of the distance between successive markers associated with a catheter the accuracy of a catheter shift measurement was equal to 5%. Visual inspection proved that the observed measurement errors were associated with faster movements of a catheter. Bronchoscope redesign option is proposed to improve catheter shift measurement accuracy. The results of this study demonstrate that application of image analysis techniques to data recorded during bronchoscopy examination can at least support the existing navigation methods for peripheral bronchoscopy with respect to the determination of the location of the catheter distal tip within the lumen of the pulmonary airways.

7

Cases of application of DVR for reconstruction of road accidents

Abramowski M. M.

Zeszyty Naukowe Instytutu Pojazdów / Politechnika Warszawska

|

2016

|

z. 5/109

5--15

EN

The article presents the theoretical assumptions of the analysis of images recorded by the video recorder in an accident. Methods shown in work [3] and in this article describe how to use the video recorded by the DVR (as well as surveillance traffic cameras) you can analyze and determine a lot of valuable information to facilitate the reconstruction of road accidents. The idea is to build a tool to support an expert in accident reconstruction, to be used in the reconstruction of recordings from recording devices.

PL

W artykule przedstawiono założenia teoretyczne analizy obrazu zarejestrowanego przez wideorejestrator podczas wypadku drogowego. Metody pokazane w pracy [3] jak i w tym artykule przedstawiają sposób w jaki można wykorzystać materiał filmowy zapisany przez wideorejestrator (a także kamery monitorujące ruch drogowy) można przeanalizować i wyznaczyć wiele cennych danych ułatwiających rekonstrukcję zdarzeń drogowych. Założeniem jest zbudowanie narzędzia wspomagającego biegłego w rekonstrukcji wypadków drogowych z możliwością wykorzystania w tej rekonstrukcji nagrań z urządzeń rejestrujących.

8

Ocena bezpieczeństwa na przejściach dla pieszych przy pomocy analizy obrazu wideo

Szagała P., Czajewski W., Dąbkowski P., Olszewski P.

Czasopismo Inżynierii Lądowej, Środowiska i Architektury / Journal of Civil Engineering, Environment and Architecture

|

2016

|

z. 63, nr 1/II

331--341

PL

Pomimo ogólnej poprawy bezpieczeństwa ruchu drogowego w Polsce, liczba wypadków na przejściach dla pieszych nie zmniejszyła się w ciągu ostatnich czterech lat. W niniejszym artykule przedstawiono rezultaty projektu badawczego MOBIS, którego celem było opracowanie zastępczych miar bezpieczeństwa, opartych na wykrywaniu konfliktów pomiędzy pieszymi i pojazdami z wykorzystaniem analizy obrazu wideo. W okresie ponad 40 dni zarejestrowano ruch pieszych i pojazdów na dwóch przejściach dla pieszych bez sygnalizacji świetlnej w Warszawie i we Wrocławiu. Przy wykorzystaniu przetwarzania obrazu określono trajektorie ruchu pojazdów i pieszych. Uzyskano od 600 do 1000 sytuacji spotkań pieszy-pojazd dziennie. Obliczono parametry opisujące interakcje pieszych i pojazdów – profile prędkości, minimalną odległość między uczestnikami, przyspieszenia itp. Klasyfikacja sytuacji została oparta na interakcjach pieszych i pojazdów. Na podstawie czasu po opuszczeniu strefy konfliktu (PET) oraz prędkości pojazdu zidentyfikowano sytuacje niebezpieczne. Jako zastępczą miarę bezpieczeństwa na przejściach dla pieszych zaproponowano Wskaźnik Zagrożenia Pieszych (WZP), który określa promil sytuacji niebezpiecznych występujących na danym przejściu.

EN

Although road safety situation in Poland is generally improving, the number of accidents at pedestrian crossings has not decreased in the last four years. The paper presents results of research project MOBIS, the aim of which was to develop surrogate safety indicators, based on detection of pedestrian-vehicle conflicts using video analysis. Pedestrian and vehicle traffic was filmed at two unsignalised pedestrian zebra crossings in Warsaw and Wrocław for over 40 days. Motion trajectories of vehicles and pedestrians were determined based on video processing. The number of pedestrian-vehicle encounters was between 600 and 1000 per day. Several parameters describing pedestrian-vehicle interactions were calculated, such as: speed profiles, post encroachment time, minimum distance between the participants, decelerations, etc. Classification of encounters was based on interactions of pedestrians and vehicles i.e.: driver yielding to a pedestrian, vehicle passing just in front of or behind a pedestrian. Dangerous encounters were identified based on Post-Encroachment Time (PET) as well as the vehicle speed. Dangerous Encounter Index is proposed as a surrogate safety indicator for pedestrian crossings. The index shows improvement of safety after introduction of active signage involving blinking lights at the crossings.

9

Performance evaluation of parallel background subtraction on GPU platforms

Szwoch G.

Elektronika : konstrukcje, technologie, zastosowania

|

2015

|

Vol. 56, nr 4

23-27

EN

Implementation of the background subtraction algorithm on parallel GPUs is presented. The algorithm processes video streams and extracts foreground pixels. The work focuses on optimizing parallel algorithm implementation by taking into account specific features of the GPU architecture, such as memory access, data transfers and work group organization. The algorithm is implemented in OpenCL and CUDA. Various optimizations of the algorithm are presented and tested using devices with varying processing power, including desktop PC graphic cards, ultrabooks and the Tegra mobile processor. The aim of the work is to determine if the optimized algorithm, run on currently available GPUs, is able to perform on-line processing of high resolution video streams.

PL

W artykule przedstawiono implementację algorytmu odejmowania tła na procesorach równoległych GPU. Algorytm przetwarza strumienie obrazu z kamer i wyodrębnia piksele nie należące do tła. Praca skupia się na optymalizacji równoległego przetwarzania obrazu z uwzględnieniem architektury procesorów GPU. Algorytm został zaimplementowany w systemach OpenCL i CUDA. Przedstawiono różne techniki optymalizacji i wyniki testów wykonanych na procesorach GPU w urządzeniach o różnej mocy obliczeniowej. Celem pracy jest określenie czy zoptymalizowany algorytm uruchomiony na dostępnych obecnie urządzeniach GPU jest w stanie przetwarzać strumienie obrazu w trybie online.

10

Badania naukowe w zakresie technik multimedialnych

Domański M, Bartkowiak M., Grajek T., Karwowski D., Maćkowiak S.

Przegląd Telekomunikacyjny + Wiadomości Telekomunikacyjne

|

2013

|

nr 2-3

77--80

PL

Artykuł stanowi przegląd wybranych obszarów tematycznych badań naukowych w zakresie multimediów, w których ostatnio prowadzono prace w Katedrze Telekomunikacji Multimedialnej i Mikroelektroniki Politechniki Poznańskiej. W szczególności artykuł dotyczy usprawnień adaptacyjnego kodowania arytmetycznego CABAC, badań w zakresie inteligentnych systemów nadzoru wizyjnego, wykorzystania stereoskopii do nadzoru ruchu drogowego oraz modelowania i kompresji dźwięku.

EN

The paper is a brief overview of some selected research areas treated by the team of the Chair of Multimedia Telecommunications and Microelectronics at Poznań University of Technology. The considerations of the research results of that team is limited to multimedia techniques only, while their work is also outside that discipline. In particular, the paper deals with extensions of adaptive arithmetic coding CABAC, research on intelligent video surveillance systems, applications of stereovision in road traffic surveillance and audio modeling and compression.

11

Spectral analysis by a video camera in a holographic optical tweezers setup

Drobczynski S, Dus-Szachniewicz K, Symonowicz K., Glogocka D

Optica Applicata

|

2013

|

Vol. 43, nr 4

739--746

EN

We discuss the basic parameters of the holographic tweezers equipped with a diode laser and cheap video camera. We compare these parameters with the system using a fast camera and high power Nd:YAG laser. The measured parameters are: the power spectra density calculated from tracing the position of the micron polystyrene beads and the trap stiffness. We show that this cheap optical tweezers system is sufficient for experiments in microbiology.

12

Synchronization of data recorded using acquisition stations with data from camera during the bubble departure

Dzienis P., Mosdorf R.

Advances in Science and Technology. Research Journal

|

2013

|

Vol. 7, nr 20

29--34

EN

In this study the first part of the experimental data was recorded in a data acquisition station, and another one was recorded with a high speed camera. The data recorded using the acquisition station was recorded with higher frequency than the time between two subsequent frames of the film. During the analysis of the experimental data the problem was related to the synchronization of measurement from acquisition station and data recorded with a camera. In this paper the method of synchronization of experimental data has been shown. A laser- phototransistor system has been used. The data synchronization was required in scaling of sampling frequency in the investigated time series.

13

Face Tracking and Recognition with the Use of Particle-Filtered Local Features

Stasiak Ł. A., Pacut A.

Journal of Telecommunications and Information Technology

|

2010

|

nr 4

26-36

EN

A consistent particle filtering-based framework for the purpose of parallel face tracking and recognition from video sequences is proposed. A novel approach to defining randomized, particle filtering-driven local face features for the purpose of recognition is proposed. The potential of cumulating classification decisions based on the proposed feature set definition is evaluated. By applying cumulation mechanisms to the classification results determined from single frames and with the use of particle-filtered features, good recognition rates are obtained at the minimal computational cost. The proposed framework can operate in real-time on a typical modern PC. Additionally, the application of cumulation mechanisms makes the framework resistant to brief visual distortions, such as occlusions, head rotations or face expressions. A high performance is also obtained on low resolution images (video frames). Since the framework is based on the particle filtering principle, it is easily tunable to various application requirements (security level, hardware constraints).

14

Identyfikacja człowieka na podstawie zestawu obrazów twarzy pochodzących z sekwencji video

Nowosielski A.

Metody Informatyki Stosowanej

|

2008

|

nr 1 (Tom 13)

113-125

PL

W artykule przedstawiono problem identyfikacji człowieka na podstawie zestawu obrazów twarzy pochodzących z sekwencji wideo. Zaproponowano mechanizm poszukiwania odpowiedników w lokalnym sąsiedztwie z poprzednich klatek sekwencji wideo. Wykazano, że wykorzystujący go algorytm agregacji wyników pozwala zwiększyć skuteczność identyfikacji. Końcowy wynik jest przy tym bardziej wiarygodny i stabilny. W artykule pokazano przy tym, że tradycyjny system monitoringu, który mógłby stanowić źródło sekwencji wideo do systemu rozpoznawania, niekoniecznie spełnia wymagane kryteria jakości obrazów. Zaproponowano więc odpowiednie podejścia do rozwiązania tego problemu.

EN

Traditional face recognition is based on the. static image analysis. In a real-life scenario there is a sequence of consecutive images from a video camera. By tracking a face in a sequence and aggregating individual results of face recognition, a much more stable and reliable final recognition result may be generated. Based on the analysis of the problem the algorithm for people identification from a video stream is developed.

15

Analiza natężenia ruchu osób na monitorowanych obszarach

Nowosielski A., Kłosowski K.

Metody Informatyki Stosowanej

|

2008

|

nr 4 (Tom 17)

139--150

EN

New and unfamiliar quality is being brought in with the development of intelligent monitoring systems. Video analysis of human dynamics is an excellent example. Based on the analysis of the problem the algorithm for human traffic analysis from a video stream is developed.

16

Weighted ensemble boosting for robust activity recognition in video

Ivanov Y., Hamid R.

Machine Graphics and Vision

|

2006

|

Vol. 15, No. 3/4

415-427

EN

In this paper we introduce a novel approach to classifier combination, which we term Weighted Ensemble Boosting. We apply the proposed algorithm to the problem of activity recognition in video, and compare its performance to different classifier combination methods. These include Approximate Bayesian Combination, Boosting, Feature Stacking, and the more traditional Sum and Product rules. Our proposed Weighted Ensemble Boosting algorithm combines the Bayesian averaging strategy with the boosting framework, finding useful conjunctive feature combinations and achieving a lower error rate than the traditional boosting algorithm. The method demonstrates a comparable level of stability with respect to the classifier selection pool. We show the performance of our technique for a set of 6 types of classifiers in an office setting, detecting 7 classes of typical office activities.