An efficient and robust system for multi-person event detection in real world indoor surveillance scenes

Due to the popularity of security cameras in public places, it is of interest to design an intelligent system that can efficiently detect events automatically. This paper proposes a novel algorithm for multi-person event detection. To ensure greater than real-time performance, features are extracted directly from compressed MPEG video. A novel histogram-based feature descriptor that captures the angles between extracted particle trajectories is proposed, which allows us to capture motion patterns of multi-person events in the video. To alleviate the need for fine-grained annotation, we propose the use of Labelled Latent Dirichlet Allocation, a “weakly supervised” method that allows the use of coarse temporal annotations which are much simpler to obtain. This novel system is able to run at approximately ten times real-time, while preserving state-of-theart detection performance for multi-person events on a 100-hour real-world surveillance dataset (TRECVid SED).

Hierarchical semi-markov conditional random fields for recursive sequential data

Inspired by the hierarchical hidden Markov models (HHMM), we present the hierarchical semi-Markov conditional random field (HSCRF), a generalisation of embedded undirected Markov chains to model complex hierarchical, nested Markov processes. It is parameterised in a discriminative framework and has polynomial time algorithms for learning and inference. Importantly, we develop efficient algorithms for learning and constrained inference in a partially-supervised setting, which is important issue in practice where labels can only be obtained sparsely. We demonstrate the HSCRF in two applications: (i) recognising human activities of daily living (ADLs) from indoor surveillance cameras, and (ii) noun-phrase chunking. We show that the HSCRF is capable of learning rich hierarchical models with reasonable accuracy in both fully and partially observed data cases.

RAPID HUMAN ACTION RECOGNITION IN H.264/AVC COMPRESSED DOMAIN FOR VIDEO SURVEILLANCE

This paper discusses a novel high-speed approach for human action recognition in H. 264/AVC compressed domain. The proposed algorithm utilizes cues from quantization parameters and motion vectors extracted from the compressed video sequence for feature extraction and further classification using Support Vector Machines (SVM). The ultimate goal of our work is to portray a much faster algorithm than pixel domain counterparts, with comparable accuracy, utilizing only the sparse information from compressed video. Partial decoding rules out the complexity of full decoding, and minimizes computational load and memory usage, which can effect in reduced hardware utilization and fast recognition results. The proposed approach can handle illumination changes, scale, and appearance variations, and is robust in outdoor as well as indoor testing scenarios. We have tested our method on two benchmark action datasets and achieved more than 85% accuracy. The proposed algorithm classifies actions with speed (>2000 fps) approximately 100 times more than existing state-of-the-art pixel-domain algorithms.

Motion Detection for Video Surveillance

This thesis is related to the broad subject of automatic motion detection and analysis in videosurveillance image sequence. Besides, proposing the new unique solution, some of the previousalgorithms are evaluated, where some of the approaches are noticeably complementary sometimes.In real time surveillance, detecting and tracking multiple objects and monitoring their activities inboth outdoor and indoor environment are challenging task for the video surveillance system. Inpresence of a good number of real time problems limits scope for this work since the beginning. Theproblems are namely, illumination changes, moving background and shadow detection.An improved background subtraction method has been followed by foreground segmentation, dataevaluation, shadow detection in the scene and finally the motion detection method. The algorithm isapplied on to a number of practical problems to observe whether it leads us to the expected solution.Several experiments are done under different challenging problem environment. Test result showsthat under most of the problematic environment, the proposed algorithm shows the better qualityresult.

Multiple camera coordination in a surveillance system

We present a distributed surveillance system that uses multiple cheap static cameras to track multiple people in indoor environments. The system has a set of Camera Processing Modules and a Central Module to coordinate the tracking tasks among the cameras. Since each object in the scene can be tracked by a number of cameras, the problem is how to choose the most appropriate camera for each object. This is important given the need to deal with limited resources (CPU, power etc.). We propose a novel algorithm to allocate objects to cameras using the object-to-camera distance while taking into account occlusion. The algorithm attempts to assign objects in the overlapping field of views to the nearest camera, which can see the object without occlusion. Experimental results show that the system can coordinate cameras to track people and can deal well with occlusion.

Combining image regions and human activity for indirect object recognition in indoor wide-angle views

Traditional methods of object recognition are reliant on shape and so are very difficult to apply in cluttered, wideangle and low-detail views such as surveillance scenes. To address this, a method of indirect object recognition is proposed, where human activity is used to infer both the location and identity of objects. No shape analysis is necessary. The concept is dubbed 'interaction signatures', since the premise is that a human will interact with objects in ways characteristic of the function of that object - for example, a person sits in a chair and drinks from a cup. The human-centred approach means that recognition is possible in low-detail views and is largely invariant to the shape of objects within the same functional class. This paper implements a Bayesian network for classifying region patches with object labels, building upon our previous work in automatically segmenting and recognising a human's interactions with the objects. Experiments show that interaction signatures can successfully find and label objects in low-detail views and are equally effective at recognising test objects that differ markedly in appearance from the training objects.

Foreground segmentation in depth imagery using depth and spatial dynamic models for video surveillance applications

Low-cost systems that can obtain a high-quality foreground segmentation almostindependently of the existing illumination conditions for indoor environments are verydesirable, especially for security and surveillance applications. In this paper, a novelforeground segmentation algorithm that uses only a Kinect depth sensor is proposedto satisfy the aforementioned system characteristics. This is achieved by combininga mixture of Gaussians-based background subtraction algorithm with a new Bayesiannetwork that robustly predicts the foreground/background regions between consecutivetime steps. The Bayesian network explicitly exploits the intrinsic characteristics ofthe depth data by means of two dynamic models that estimate the spatial and depthevolution of the foreground/background regions. The most remarkable contribution is thedepth-based dynamic model that predicts the changes in the foreground depth distributionbetween consecutive time steps. This is a key difference with regard to visible imagery,where the color/gray distribution of the foreground is typically assumed to be constant.Experiments carried out on two different depth-based databases demonstrate that theproposed combination of algorithms is able to obtain a more accurate segmentation of theforeground/background than other state-of-the art approaches.

Characterisation of indoor air organic pollutants in Brisbane

The occurrence and levels of airborne polycyclic aromatic hydrocarbons and volatile organic compounds in selected non-industrial environments in Brisbane have been investigated as part of an integrated indoor air quality assessment program. The most abundant and most frequently encountered compounds include, nonanal, decanal, texanol, phenol, 2-ethyl-1-hexanol, ethanal, naphthalene, 2,6-tert-butyl-4-methyl-phenol (BHT), salicylaldehyde, toluene, hexanal, benzaldehyde, styrene, ethyl benzene, o-, m- and pxylenes, benzene, n-butanol, 1,2-propandiol, and n-butylacetate. Many of the 64 compounds usually included in the European Collaborative Action method of TVOC analysis were below detection limits in the samples analysed. In order to extract maximum amount of information from the data collected, multivariate data projection methods have been employed. The implications of the information extracted on source identification and exposure control are discussed.

A pilot study on the effect of indoor particle sources on indoor particle concentration in residential houses

Characterization of indoor particle sources from 14 residential houses in Brisbane, Australia, was performed. The approximation of PM2.5 and the submicrometre particle number concentrations were measured simultaneously for more than 48 h in the kitchen of all the houses by using a photometer (DustTrak) and a condensation particle counter (CPC), respectively. From the real time indoor particle concentration data and a diary of indoor activities, the indoor particle sources were identified. The study found that among the indoor activities recorded in this study, frying, grilling, stove use, toasting, cooking pizza, smoking, candle vaporizing eucalyptus oil and fan heater use, could elevate the indoor particle number concentration levels by more than five times. The indoor approximation of PM2.5 concentrations could be close to 90 times, 30 times and three times higher than the background levels during grilling, frying and smoking, respectively.