Outlier detection has recently become an important problem in many industrial and financial applications. Often, outliers have to be detected from data streams that continuously arrive from data sources. Incremental outlier detection algorithms, aimed at detecting outliers as soon as they appear in a database, have recently become emerging research field. In this paper, we develop an incremental version of connectivity-based outlier factor (COF) algorithm and discuss its computational complexity. The proposed incremental COF algorithm has equivalent detection performance as the iterated static COF algorithm (applied after insertion of each data record), with significant reduction in computational time. The paper provides theoretical and experimental evidence that the number of updates per such insertion/deletion does not depend on the total number of points in the data set, which makes algorithm viable for very large dynamic datasets. Finally, we also illustrate an application of the proposed algorithm on motion detection in video surveillance applications.

Recent research in motion detection has shown that various outlier detection methods could be used for efficient detection of small moving targets. These algorithms detect moving objects as outliers in a properly defined attribute space, where outlier is defined as an object distinct from the objects in its neighborhood. In this paper, we compare the performance of two incremental outlier detection algorithms, namely the incremental connectivity-based outlier factor and the incremental local outlier factor to modified Stauffer-Grimson algorithm. Each video sequence is represented with spatial-temporal blocks extracted from the raw video. Principal component analysis (PCA) is applied on these blocks in order to reduce the dimensionality of extracted data. Extensive experiments performed on several data sets, including infrared sequences from OSU Thermal Pedestrian Database repository, and data collected at Delaware State University from FLIR Systems PTZ cameras have shown promising results in using outlier detection for detection of small moving targets.

In this paper, we describe a technique for detection of moving objects in RGB and infra-red (IR) videos. The technique is based on novel incremental connectivity-based outlier factor (IncCOF). The main idea of the proposed approach is to detect moving blocks as outliers-objects dissimilar to objects in their vicinity-within a properly defined feature space. As the feature space, we use representation of videos by spatial-temporal blocks combined with principal component analysis for dimensionality reduction. Experimental evaluation of the proposed approach on a variety of test videos, including PETS repository, demonstrates its applicability and robustness on the choice of parameters.

No abstract