A Spatial-Temporal Feature Extraction and Unsupervised Recognition of Video Anomalies in Real Time

Anomaly detection is an area of video analysis has a great importance in automated surveillance. Although it has been extensively studied, there has been little work started using CNN networks. Hence, in this thesis we presented a novel approach for learning motion features and modeling normal Spatio-temporal dynamics for anomaly detection. In our technique, we capture variations in scale of the patterns of motion in an image object by using optical flow dense estimation technique and train our auto encoder model using convolution long short term memories (ConvLSTM2D) as we are processing video frames and we predict the anomaly in real time using Euclidean distance between the generated and the ground truth frame and we achieved a real time accuracy of nearly 98% for the youtube videos which are not used for either testing or training. Error between the network’s output and the target output is used to classify a video volume as normal or abnormal. In addition to the use of reconstruction error, we also use prediction error for anomaly detection. The prediction models show comparable performance with state of the art methods. In comparison with the proposed method, performance is improved in one dataset. Moreover, running time is significantly faster.

Download Full-text

Real-time characterization of the spatio-temporal dynamics of deformable mirrors

10.1117/12.2238431 ◽

2016 ◽

Author(s):

James Kilpatrick ◽

Adela Apostol ◽

Anatoliy Khizhnya ◽

Vladimir Markov ◽

Leonid Beresnev

Keyword(s):

Real Time ◽

Temporal Dynamics ◽

Deformable Mirrors ◽

Spatio Temporal

Download Full-text

A Hybrid Deep Learning Framework for Unsupervised Anomaly Detection in Multivariate Spatio-Temporal Data

Applied Sciences ◽

10.3390/app10155191 ◽

2020 ◽

Vol 10 (15) ◽

pp. 5191

Author(s):

Yıldız Karadayı ◽

Mehmet N. Aydin ◽

A. Selçuk Öğrenci

Keyword(s):

Deep Learning ◽

Anomaly Detection ◽

Hurricane Katrina ◽

Time Series Data ◽

Hybrid Approach ◽

Ground Truth ◽

Outbreak Detection ◽

Series Data ◽

Detection Techniques ◽

Spatio Temporal

Multivariate time-series data with a contextual spatial attribute have extensive use for finding anomalous patterns in a wide variety of application domains such as earth science, hurricane tracking, fraud, and disease outbreak detection. In most settings, spatial context is often expressed in terms of ZIP code or region coordinates such as latitude and longitude. However, traditional anomaly detection techniques cannot handle more than one contextual attribute in a unified way. In this paper, a new hybrid approach based on deep learning is proposed to solve the anomaly detection problem in multivariate spatio-temporal dataset. It works under the assumption that no prior knowledge about the dataset and anomalies are available. The architecture of the proposed hybrid framework is based on an autoencoder scheme, and it is more efficient in extracting features from the spatio-temporal multivariate datasets compared to the traditional spatio-temporal anomaly detection techniques. We conducted extensive experiments using buoy data of 2005 from National Data Buoy Center and Hurricane Katrina as ground truth. Experiments demonstrate that the proposed model achieves more than 10% improvement in accuracy over the methods used in the comparison where our model jointly processes the spatial and temporal dimensions of the contextual data to extract features for anomaly detection.

Download Full-text

Spatio-temporal texture modelling for real-time crowd anomaly detection

Computer Vision and Image Understanding ◽

10.1016/j.cviu.2015.08.010 ◽

2016 ◽

Vol 144 ◽

pp. 177-187 ◽

Cited By ~ 25

Author(s):

Jing Wang ◽

Zhijie Xu

Keyword(s):

Anomaly Detection ◽

Real Time ◽

Spatio Temporal ◽

Texture Modelling ◽

Crowd Anomaly Detection

Download Full-text

Spatio-temporal multidimensional collective data analysis for providing comfortable living anytime and anywhere

APSIPA Transactions on Signal and Information Processing ◽

10.1017/atsip.2018.4 ◽

2018 ◽

Vol 7 ◽

Cited By ~ 2

Author(s):

Naonori Ueda ◽

Futoshi Naya

Keyword(s):

Data Analysis ◽

Real Time ◽

Urban Areas ◽

Temporal Data ◽

Collective Data ◽

Novel Approach ◽

Smart Data ◽

Spatio Temporal ◽

World Information ◽

Temporal Data Analysis

Machine learning is a promising technology for analyzing diverse types of big data. The Internet of Things era will feature the collection of real-world information linked to time and space (location) from all sorts of sensors. In this paper, we discuss spatio-temporal multidimensional collective data analysis to create innovative services from such spatio-temporal data and describe the core technologies for the analysis. We describe core technologies about smart data collection and spatio-temporal data analysis and prediction as well as a novel approach for real-time, proactive navigation in crowded environments such as event spaces and urban areas. Our challenge is to develop a real-time navigation system that enables movements of entire groups to be efficiently guided without causing congestion by making near-future predictions of people flow. We show the effectiveness of our navigation approach by computer simulation using artificial people-flow data.

Download Full-text

A hybrid SIR model applied to “Covid- 19” pandemic

10.21203/rs.3.rs-83509/v1 ◽

2020 ◽

Author(s):

samuel Ismael BILLONG IV ◽

Georges Edouard Kouamou ◽

Thomas Bouetou

Keyword(s):

Prediction Models ◽

Temporal Dynamics ◽

Equation System ◽

Sir Model ◽

Diffusion Equations ◽

Statistical Prediction ◽

Accurate Information ◽

Test Case ◽

Developed Country ◽

Spatio Temporal

Abstract Modeling has become a tool capable of guiding public policies, especially in the area of health. Specifically, modeling in epidemiology makes it possible to follow the evolution of infections and to understand the behavior of viruses. Unfortunately, the traditional SIR models and the statistical prediction models commonly used suffer from the lack of accurate information and the unavailability of a large amount of data, also they do not take into account the interactions within the population. This paper proposes a hybrid SIR model which takes into consideration the spatio-temporal dynamics of individuals. The model is based on the discrete stochastic diffusion equations. To build the equation system, the 2D diffusion equations are coupled to the human displacement probability law pattern through a discretization made by the finite volume method for complex geometries. Beyond the health consequences it causes, COVID-19 (coronavirus) is a test case used to validate the proposed model. We used the case of a developed country before confinement to fit to the chosen displacement pattern, and to analyze the sensitivity of the parameters of the model taking into account the accuracy of the statistics provided.

Download Full-text

A novel localization based biosensor, SEDAR, visualizes spatio-temporal dynamics of activated Ras in endogenous Drosophila tissues

10.1101/432997 ◽

2018 ◽

Author(s):

Rishita Changede

Keyword(s):

Plasma Membrane ◽

Growth Factor ◽

Real Time ◽

Tyrosine Kinases ◽

Temporal Dynamics ◽

Three Dimensional ◽

Growth Factor Receptor ◽

Leading Edge ◽

Border Cells ◽

Spatio Temporal

AbstractChemokine signaling via growth factor receptor tyrosine kinases (RTKs) regulates development, differentiation, growth and disease implying that it is involved in a myriad of cellular processes. A single RTK, for example the Epidermal Growth Factor Receptor (EGFR), is used repeatedly for a multitude of developmental programs. Quantitative differences in magnitude and duration of RTK signaling can bring about different signaling outcomes. Understanding this complex RTK signals requires real time visualization of the signal. To visualize spatio-temporal signaling dynamics, a biosensor called SEnsitive Detection of Activated Ras (SEDAR) was developed. It is a localization-based sensor that binds to activated Ras directly downstream of the endogenous activated RTKs. This binding was reversible and SEDAR expression did not cause any detectable perturbation of the endogenous signal. Using SEDAR, endogenous guidance signaling was visualized during RTK mediated chemotaxis of border cells in Drosophila ovary. SEDAR localized to both the leading and rear end of the cluster but polarized at the leading edge of the cluster. Perturbation of RTKs that led to delays in forward migration of the cluster correlated with loss of SEDAR polarization in the cluster. Gliding or tumbling behavior of border cells was a directly related to the high or low magnitude of SEDAR polarization respectively, in the leading cell showing that signal polarization at the plasma membrane provided information for the migratory behavior. Further, SEDAR localization to the plasma membrane detected EGFR mediated signaling in five distinct developmental contexts. Hence SEDAR, a novel biosensor could be used as a valuable tool to study the dynamics of endogenous Ras activation in real time downstream of RTKs, in three-dimensional tissues, at an unprecedented spatial and temporal resolution.

Download Full-text

Mobility Prediction-Based Optimisation and Encryption of Passenger Traffic-Flows Using Machine Learning

Sensors ◽

10.3390/s20092629 ◽

2020 ◽

Vol 20 (9) ◽

pp. 2629

Author(s):

Syed Muhammad Asad ◽

Jawad Ahmad ◽

Sajjad Hussain ◽

Ahmed Zoha ◽

Qammer Hussain Abbasi ◽

...

Keyword(s):

Machine Learning ◽

Real Time ◽

Radio Frequency Identification ◽

City Planning ◽

Prediction Models ◽

Mobility Prediction ◽

Traffic Flows ◽

Passenger Traffic ◽

Passenger Flow ◽

Spatio Temporal

Information and Communication Technology (ICT) enabled optimisation of train’s passenger traffic flows is a key consideration of transportation under Smart City planning (SCP). Traditional mobility prediction based optimisation and encryption approaches are reactive in nature; however, Artificial Intelligence (AI) driven proactive solutions are required for near real-time optimisation. Leveraging the historical passenger data recorded via Radio Frequency Identification (RFID) sensors installed at the train stations, mobility prediction models can be developed to support and improve the railway operational performance vis-a-vis 5G and beyond. In this paper we have analysed the passenger traffic flows based on an Access, Egress and Interchange (AEI) framework to support train infrastructure against congestion, accidents, overloading carriages and maintenance. This paper predominantly focuses on developing passenger flow predictions using Machine Learning (ML) along with a novel encryption model that is capable of handling the heavy passenger traffic flow in real-time. We have compared and reported the performance of various ML driven flow prediction models using real-world passenger flow data obtained from London Underground and Overground (LUO). Extensive spatio-temporal simulations leveraging realistic mobility prediction models show that an AEI framework can achieve 91.17% prediction accuracy along with secure and light-weight encryption capabilities. Security parameters such as correlation coefficient (<0.01), entropy (>7.70), number of pixel change rate (>99%), unified average change intensity (>33), contrast (>10), homogeneity (<0.3) and energy (<0.01) prove the efficacy of the proposed encryption scheme.

Download Full-text