Data exploitation using visual analytics

In a surveillance system the huge volume of recorded multidimensional data poses a great challenge to the user in performing meaningful analysis in efficient and coherent manner, especially in a human-vehicle or human-object interaction domain. To address this concern a semi automated data analysis concept is developed for feature extraction, object detection, trajectory determination and cluster identification. In addition this paper presents an algorithmic basis for significantly improved correlation and association of features, and events of interest in a timely and sound manner. The uncertainty associated with the operator's interpretation of data is tackled by proposing an acceptable hypothesis by the analyst based on human intelligence and experience. Experimental results and graphs are also presented in this paper.