Abnormality Detection using K-means Data Stream Clustering Algorithm in Intelligent Surveillance System

Abstract

In this research work a k-Means clustering technique utilized in a new data stream clustering method used in abnormal detection system. This system implies the use of a set of features (such as: distance, direction, x-coordinate, y-coordinate) extracted from set of pairs of interest point that obtained using HARRIS or FAST detector from the frames of video clips in two publically available datasets, the first UCSD pedestrian dataset (ped1 and ped2 datasets), and the second VIRAT video dataset. The results indicated that using HARRIS detector achieved detection rate 1% with 6% false alarms by using UCSD (Ped1) dataset, 10.75% detection Rate with 10% false alarm rate by using UCSD (Ped2) dataset, and 5% detection rate with 40% false alarms by using VIRAT dataset. While for FAST detector, the achieved detection rates are 0.5%, 10.75%, and 4.08% while the false alarm rates are 5%, 10.50%, and 45.92% by using UCSD (Ped1), UCSD (Ped2), and VIRAT datasets respectively.