Olga Isupova

Learning Methods for Dynamic Topic Modeling in Automated Behavior Analysis

Semisupervised and unsupervised systems provide operators with invaluable support and can tremendously reduce the operators’ load. In the light of the necessity to process large volumes of video data and provide autonomous decisions, this paper proposes new learning algorithms for activity analysis in video. The activities and behaviors are described by a dynamic topic model. Two novel learning algorithms based on the expectation maximization approach and variational Bayes inference are proposed. Theoretical derivations of the posterior estimates of model parameters are given. The designed learning algorithms are compared with the Gibbs sampling inference scheme introduced earlier in the literature. A detailed comparison of the learning algorithms is presented on real video data. We also propose an anomaly localization procedure, elegantly embedded in the topic modeling framework. It is shown that the developed learning algorithms can achieve 95% success rate. The proposed framework can be applied to a number of areas, including transportation systems, security, and surveillance.

Proposed Learning Algorithms for Markov Clustering Topic Model

This chapter introduces the methods for the behaviour analysis and anomaly detection in video using a topic model. Topics in video applications represent typical motion patterns in an observed scene. These patterns can be used for semantic understanding of the typical activities happening within the scene. They can also be used to detect abnormal events. Likelihood of newly observed data is employed as a measure of normality. If something atypical happens in a new visual document, then this document cannot be fitted with the topics, or typical activities, learnt before, and it would have a low likelihood value. The focus of this chapter is on development and comparison of learning algorithms for the Markov clustering topic model. A novel anomaly localisation procedure is also introduced in this chapter.

Change Point Detection with Gaussian Processes

This chapter introduces a novel framework for detecting anomalies as change points. This chapter presents a general approach for change point detection, which can be used for behaviour analysis (where periods between change points are considered as different behaviours) and anomaly detection (where a change is considered as a break point between normal and abnormal behaviours). In the proposed framework changes are considered as functional breaks in input data.

Dynamic Hierarchical Dirichlet Process

This chapter introduces a novel dynamic nonparametric topic model that allows a potentially infinite number of topics and, in practice, the number of topics is determined by the data. The application of the model for behaviour analysis and anomaly detection in video is considered in detail, however, the proposed model is not limited to this application.

Online vehicle logo recognition using Cauchy prior logistic regression

Vehicle logo recognition is an important part of vehicle identification in intelligent transportation systems. State-of-the-art vehicle logo recognition approaches typically consider training models on large datasets. However, there might only be a small training dataset to start with and more images can be obtained during the real-time applications. This paper proposes an online image recognition framework which provides solutions for both small and large datasets. Using this recognition framework, models are built efficiently using a weight updating scheme. Another novelty of this work is that the Cauchy prior logistic regression with conjugate gradient descent is proposed to deal with the multinomial classification tasks. The Cauchy prior results in a quicker convergence speed for the weight updating process which could decrease the computational cost for both online and offline methods. By testing with a publicly available dataset, the Cauchy prior logistic regression deceases the classification time by 59%. An accuracy of up to 98.80% is achieved when the proposed framework is applied.

Autonomous flame detection in video based on saliency analysis and optical flow

The paper proposes a flame detection method based on saliency analysis, optical flow estimation and temporal wavelet transform. Two separate saliency maps are first obtained based on the grayscale values and optical flow magnitudes of each frame using a saliency detector. Subsequently, the two maps are combined to extract candidate flame regions. To further discard falsely detected pixels, a colour model of flames and temporal wavelet transform are employed. The proposed algorithms can be applied in the autonomous and semi-autonomous systems for environmental surveillance and can reduce the load of human operators. Experiments illustrate the introduced method achieves around 91% true positive rate and 97% true negative rate.

Compressive sensing approaches for autonomous object detection in video sequences

Video analytics requires operating with large amounts of data. Compressive sensing allows to reduce the number of measurements required to represent the video using the prior knowledge of sparsity of the original signal, but it imposes certain conditions on the design matrix. The Bayesian compressive sensing approach relaxes the limitations of the conventional approach using the probabilistic reasoning and allows to include different prior knowledge about the signal structure. This paper presents two Bayesian compressive sensing methods for autonomous object detection in a video sequence from a static camera. Their performance is compared on real datasets with the non-Bayesian greedy algorithm. It is shown that the Bayesian methods can provide more effective results than the greedy algorithm in terms of both accuracy and computational time.