An entropy approach for abnormal activities detection in video streams

Detection of aberration in video surveillance is an important task for public safety. This paper puts forward a simple but effective framework to detect aberrations in video streams using Entropy, which is estimated on the statistical treatments of the spatiotemporal information of a set of interest points within a region of interest by measuring their degree of randomness of both directions and displacements. Entropy is a measure of the disorder/randomness in video frame. It has been showed that degree of randomness of the directions (circular variance) changes markedly in abnormal state of affairs and does change only direction variation but does not change with displacement variation of the interest point. Degree of randomness of the displacements has been put in for to counterbalance this deficiency. Simple simulations have been exercised to see the characteristics of these crude elements of entropy. Normalized entropy measure provides the knowledge of the state of anomalousness. Experiments have been conducted on various real world video datasets. Both simulation and experimental results report that entropy measures of the frames over time is an outstanding way to characterize anomalies in videos.