Analyzing the Performance of DTBO in Single-View Surveillance Video Synopsis

Nowadays, surveillance cameras are ruling the world in terms of security measures. Worldwide application of surveillance cameras leads to the production of huge amounts of video data. Investigating and getting the necessary information from such a lengthy video is a challenging task. To ease the inspection, Video Synopsis (VS) technology was proposed. Video Synopsis is the technique used to shorten a long-duration video with the preservation of all activities that have occurred originally. The synopsis of a video can be produced through four modules, namely, Object Detection, Object Tracking, Optimization, and finally Stitching. The efficient implementation of the VS is mainly dependent on the Optimization module. Mostly, Simulated Annealing (SA) is applied to solve the optimization problem in Object-based Single-camera Offline Surveillance Video Synopsis. In this paper, an attempt is made to implement and study the performance of the Driving Training Based Optimization (DTBO) algorithm for solving the optimization problem. The experimental results of DTBO are compared with the performance of the SA, Non-dominated Sorting Genetic Algorithm II (NSGA-II), Forest Optimization Algorithm (FOA), Elltist-Jaya, and SAMP-Jaya Algorithm. From the results and comparison, it is observed that the effectiveness of the DTBO algorithm in minimizing the energy function of Object-based Single-camera Offline Surveillance Video Synopsis is superior to the existing algorithms.