We propose a novel local nearest neighbor distance (LNND) descriptor for anomaly detection in crowded scenes. Comparing with the commonly used low-level feature descriptors in previous works, LNND descriptor has two major advantages. First, LNND descriptor efficiently incorporates spatial and temporal contextual information around the video event that is important for detecting anomalous intera...

It is difficult to recognize an object that falls in the peripheral visual field; it is even more difficult when there are other objects surrounding it. This effect, known as crowding, could be due to interactions between the low-level parts or features of the surrounding objects. Here, we investigated whether crowding can also occur selectively between higher level object representations. Many...

Virtual training environments are an important tool for training military personnel in a costeffective way, and realistic crowd modeling and population simulation are crucial components for developing training scenarios set in urban environments. The key objectives of this project were to: 1. create computational models of human group behavior; 2. analyze group behavior in simulation environmen...

In this paper, we propose a novel framework for abnormal event detection in crowded scenes. A new concept of atomic event is introduced into this framework, which is the basic component of video events. Different from previous bag-of-words (BoW) modeling-based methods that represent feature descriptors using only one code word, a feature descriptor is represented using a few more atomic events ...

This paper presents two novel approaches for people counting in crowded and open environments that combine the information gathered by multiple views. Multiple camera are used to expand the field of view as well as to mitigate the problem of occlusion that commonly affects the performance of counting methods using single cameras. The first approach is regarded as a direct approach and it attemp...

The worldwide increasing sentiment of insecurity gave birth to a new era, shaking thereby the intelligent videosurveillance systems design and deployment. The large-scale use of these means has prompted the creation of new needs in terms of analysis and interpretation. For this purpose, behavior recognition and scene understanding related applications have become more captivating to a significa...

Depending on the context of interest, an anomaly is defined differently. In case when a video event isn't expected to take place in video, it seen as anomaly. It can be difficult describe uncommon events complicated scenes, but this problem frequently resolved by using high-dimensional features well descriptors. There difficulty creating reliable model trained with these descriptors because nee...

Person re-identification (PRID) has been thoroughly researched in typical surveillance scenarios where various scenes are monitored by side-mounted, rectilinear-lens cameras. To date, few methods have proposed for fisheye cameras mounted overhead and their performance is lacking. In order to close this gap, we propose a multi-feature framework PRID combine deep-learning, color-based location-ba...

In this thesis, I investigate how scenes are represented by the human visual system and how observers use visual information to reorient themselves within a space. Scenes, like objects, are three-dimensional spaces that are experienced through twodimensional views and must be recognized from many different angles. Just as people show a preference for canonical views of objects, which best show ...

Because the environment is cluttered, objects rarely appear in isolation. The visual system must therefore attentionally select behaviorally relevant objects from among many irrelevant ones. A limit on our ability to select individual objects is revealed by the phenomenon of visual crowding: an object seen in the periphery, easily recognized in isolation, can become impossible to identify when ...