Search results

We describe an intelligent video categorization engine (IVCE) that uses the learning capability of artificial neural networks (ANNs) to classify suitably preprocessed video segments into a predefined number of semantically meaningful events (categories).

We provide a survey of existing techniques that have been proposed, either directly or indirectly, towards achieving intelligent video categorization. We also compare the performance of two popular ANNs: Kohonen's self‐organizing map (SOM) and fuzzy adaptive resonance theory (Fuzzy ART). In particular, the ANNs are trained offline to form the necessary knowledge base prior to online categorization.

Experimental results show that accurate categorization can be achieved near instantaneously.

The main limitation of this research is the need for a finite set of predefined categories. Further research should focus on generalization of such techniques.

Machine understanding of video footage has tremendous potential for three reasons. First, it enables interactive broadcast of video. Second, it allows unequal error protection for different video shots/segments during transmission to make better use of limited channel resources. Third, it provides intuitive indexing and retrieval for video‐on‐demand applications.

WeChat is the largest acquaintance social networking platform in China, in which users can view and reshare web pages shared by friends. This paper aims to analyze the spatio-temporal dynamics of web pages diffused in WeChat and advice on commercials.

A large number of web pages diffused in WeChat are collected and exclusively divided into four categories according to their titles, including advertisements, news bulletins, holiday greetings and emotional essays. For each web page, an information cascade (tree structure) is constructed to describe the diffusion trace. Based on the categories, the spatio-temporal popularity is characterized; the topological, temporal and spatial properties are examined; and the spatio-temporal diffusion velocity is explored.

Through comparative analysis, different categories of pages show diversity. For spatio-temporal popularity, there is no significant difference in cascade size; holiday greetings usually last for a relatively short time on average; emotional essays are more likely to spread to more provinces. For topological, temporal and spatial characteristics, the diffusion process of advertisements is more likely to be broadcasting than other categories; news bulletins and holiday greetings have an obvious bursty; the number of viewing behavior decreases from east to west in general. For spatio-temporal diffusion velocity, emotional essays diffuse the fastest in topological and spatio-temporal dimensions.

These findings contribute to promoting products and providing support for data driven modeling of information diffusion and human activity in spatio-temporal dimensions.

The purpose of this paper is to develop a real-time trajectory planner with optimal maneuver for autonomous vehicles to deal with dynamic obstacles during parallel parking.

To deal with dynamic obstacles for autonomous vehicles during parking, a long- and short-term mixed trajectory planning algorithm is proposed in this paper. In long term, considering obstacle behavior, A-star algorithm was improved by RS curve and potential function via spatio-temporal map to obtain a safe and efficient initial trajectory. In short term, this paper proposes a nonlinear model predictive control trajectory optimizer to smooth and adjust the trajectory online based on the vehicle kinematic model. Moreover, the proposed method is simulated and verified in four common dynamic parking scenarios by ACADO Toolkit and QPOASE solver.

Compared with the spline optimization method, the results show that the proposed method can generate efficient obstacle avoidance strategies, safe parking trajectories and control parameters such as the front wheel angle and velocity in high-efficient central processing units.

It is aimed at improving the robustness of automatic parking system and providing a reference for decision-making in a dynamic environment.

This study aims to use opportunity as a theoretical lens to investigate how the spatio-temporal and social dimensions of the consumption environment create perceived opportunities for consumers to misbehave.

Drawing on routine activity theory and social impact theory, the authors use two experiments to demonstrate that spatio-temporal and social dimensions can explain consumer theft in retail settings.

Study 1 reveals mixed empirical support for the basic dimensions of routine activity theory, which posits that the opportunity to thieve is optimised when a motivated offender, suitable target and the absence of a capable formal guardian transpire in time and space. Extending the notion of guardianship, Study 2 tests social impact theory and shows that informal guardianship impacts the likelihood of theft under optimal routine activity conditions.

The study findings highlight important implications for academicians and retail managers: rather than focusing on the uncontrollable characteristics of thieving offenders, more controllable spatio-temporal and social factors of the retail environment can be actively monitored and manipulated to reduce perceived opportunities for consumer misbehaviour.

This paper aims to construct a spatio-temporal emotional framework (STEF) for digital humanities from a quantitative perspective, applying knowledge extraction and mining technology to promote innovation of humanities research paradigm and method.

The proposed STEF uses methods of information extraction, sentiment analysis and geographic information system to achieve knowledge extraction and mining. STEF integrates time, space and emotional elements to visualize the spatial and temporal evolution of emotions, which thus enriches the analytical paradigm in digital humanities.

The case study shows that STEF can effectively extract knowledge from unstructured texts in the field of Chinese Qing Dynasty novels. First, STEF introduces the knowledge extraction tools – MARKUS and DocuSky – to profile character entities and perform plots extraction. Second, STEF extracts the characters' emotional evolutionary trajectory from the temporal and spatial perspective. Finally, the study draws a spatio-temporal emotional path figure of the leading characters and integrates the corresponding plots to analyze the causes of emotion fluctuations.

The STEF is constructed based on the “spatio-temporal narrative theory” and “emotional narrative theory”. It is the first framework to integrate elements of time, space and emotion to analyze the emotional evolution trajectories of characters in novels. The execuability and operability of the framework is also verified with a case novel to suggest a new path for quantitative analysis of other novels.

The purpose of this paper is to generate information about the contours of police responsiveness, focussing on how quickly and precisely police make firearm arrests after a shooting incident.

Using a modified version of the Knox close pair method, a spatio-temporal clustering technique, over 11,000 shooting incidents and firearm arrests between 2004 and 2007 in Philadelphia, PA were analyzed.

Police are responding quickly and in a geographically targeted fashion to shootings. Across Philadelphia elevated patterns of firearm arrests were approximately two and a half times greater than would be expected if shootings and firearm arrests lacked a spatio-temporal association. Greater than expected patterns of firearm arrests persisted for roughly one-fourth of a mile and for about one week from the shooting incident but the strength of these associations waned over space and time. The pattern of police response varied slightly across different police divisions.

The current method uncovered spatio-temporal patterning and determined when these patterns were significantly different from what would be expected if the events were completely independent. Specific events and processes surrounding each event are not known.

Findings can help inform the knowledge about police behavior in terms of how police produce arrests.

The patterns observed here provide more micro-level detail than has been revealed in previous studies regarding police responsiveness to firearm violence while also introducing a more integrated spatially and temporally specific framework.

The purpose of this paper concerns the dimensions of relevance in information retrieval systems and their completeness in new retrieval contexts such as mobile search. Geography as a factor in relevance is little understood and information seeking is assumed to take place in indoor environments. Yet the rise of information seeking on the move using mobile devices implies the need to better understand the kind of situational relevance operating in this kind of context.

The paper outlines and explores a geographic information seeking process in which geographic information needs (conditioned by needs and tasks, in context) drive the acquisition and use of geographic information objects, which in turn influence geographic behaviour in the environment. Geographic relevance is defined as “a relation between a geographic information need” (like an attention span) and “the spatio‐temporal expression of the geographic information objects needed to satisfy it” (like an area of influence). Some empirical examples are given to indicate the theoretical and practical application of this work.

The paper sets out definitions of geographical information needs based on cognitive and geographic criteria, and proposes four canonical cases, which might be theorised as anomalous states of geographic knowledge (ASGK). The paper argues that geographic relevance is best defined as a spatio‐temporally extended relation between information need (an “attention” span) and geographic information object (a zone of “influence”), and it defines four domains of geographic relevance. Finally a model of geographic relevance is suggested in which attention and influence are modelled as map layers whose intersection can define the nature of the relation.

Geographic relevance is a new field of research that has so far been poorly defined and little researched. This paper sets out new principles for the study of geographic information behaviour.

This paper aims to present collection and analysis of heterogeneous urban traffic data, and integration of them through a kernel-based approach for assessing performance of urban transport network facilities. The recent development in sensing and information technology opens up opportunities for researching the use of this vast amount of new urban traffic data. This paper contributes to analysis and management of urban transport facilities.

In this paper, the data fusion algorithm are developed by using a kernel-based interpolation approach. Our objective is to reconstruct the underlying urban traffic pattern with fine spatial and temporal granularity through processing and integrating data from different sources. The fusion algorithm can work with data collected in different space-time resolution, with different level of accuracy and from different kinds of sensors. The properties and performance of the fusion algorithm is evaluated by using a virtual test bed produced by VISSIM microscopic simulation. The methodology is demonstrated through a real-world application in Central London.

The results show that the proposed algorithm is able to reconstruct accurately the underlying traffic flow pattern on transport network facilities with ordinary data sources on both virtual and real-world test beds. The data sources considered herein include loop detectors, cameras and GPS devices. The proposed data fusion algorithm does not require assumption and calibration of any underlying model. It is easy to implement and compute through advanced technique such as parallel computing.

The presented study is among the first utilizing and integrating heterogeneous urban traffic data from a major city like London. Unlike many other existing studies, the proposed method is data driven and does not require any assumption of underlying model. The formulation of the data fusion algorithm also allows it to be parallelized for large-scale applications. The study contributes to the application of Big Data analytics to infrastructure management.