ABSTRACT The rapid growth of video data has resulted in an increasing need for surveillance and violence detection systems. Although such events occur less frequently than normal activities, developing automated video surveillance systems for violence detection has become essential to minimize labor and time waste. Detecting violent activity in videos is a challenging task due to the variability and diversity of violent behavior, which can involve a wide range of actions, motions, and interactions between people and objects. Currently, researchers employ deep learning models to detect violent behaviors. In fact, a large number of deep learning approaches are based on extracting spatio‐temporal information from a video by exploiting a 3D Convolutional Neural Network (CNN). Despite their success, these techniques require a lot more parameters than 2D CNNs and have high computational complexity. Therefore, we focus on exploiting a 2D CNN to encode spatio‐temporal information. Actually, statistical features of the optical flow changes are used to give this ability to a 2D CNN. These features are designed to make attention to regions of a video clip with much more motion. Accordingly, the optical flow of an input video is calculated. To determine meaningful changes in the optical flow, the optical flow magnitude of a current frame is compared with its predecessor. After that, statistical features of these changes are extracted to summarize a video clip to a 2D template, which feeds a 2D CNN. Experimental results on four benchmark datasets observe that the suggested strategy outperforms baseline ones. In particular, we make a better estimation of the spatio‐temporal features in a video by shortening a video clip into a 2D template.

The - similarity problem, finding a group of objects which have the most similarity to each other, has become an important issue in information retrieval and data mining. The theory of this concept is mathematically proven, but it practically has high time complexity. Binary Genetic Algorithm (BGA) has been applied to improve solutions quality of this problem, but a more efficient algorithm is required. Therefore, we aim to study and compare the performance of four metaheuristic algorithms called Particle Swarm Optimization (PSO), Gravitational Search Algorithm (GSA), Imperialist Competitive Algorithm (ICA) and Fuzzy Imperialist Competitive Algorithm (FICA) to tackle this problem. The experiments are conducted on two applications; the former is on four UCI datasets as a general application and the latter is on the text resemblance application to detect multiple similar text documents from Reuters datasets as a case study. The results of experiments give a ranking of the algorithms in solving the -similarity problem in both applications based on the exploration and exploitation abilities, that the FICA achieves the first rank in both applications as well as based on the both criteria.

Gravitational search algorithm(GSA) is a recent created metaheuristic optimization algorithm with good results in function optimization as well as real world optimization problems. Many real world problems involve multiple (often conflicting) objectives, which should be optimized simultaneously. Therefore, the aim of this paper is to propose a multi-objective version of GSA, namely clustering based archive multi-objective GSA (CA-MOGSA). Proposed method is created based on the Pareto principles. Selected non-dominated solutions are stored in an external archive. To control the size of archive, the solutions with less crowding distance are removed. These strategies guarantee the elitism and diversity as two important features of multi-objective algorithms. The archive is clustered and a cluster is randomly selected for each agent to apply the gravitational force to attract it. The selection of the proper cluster is based on the distance between clusters representatives and population member (the agent). Therefore, suitable trade-off between exploration and exploitation is provided. The experimental results on eight standard benchmark functions reveal that CA-MOGSA is a well-organized multi-objective version of GSA. It is comparable with the state-of-the- art algorithms including non-dominated sorting genetic algorithm-II (NSGA-II), strength Pareto evolutionary algorithm (SPEA2) and better than multi-objective GSA (MOGSA), time-variant particle swarm optimization (TV-PSO), and non-dominated sorting GSA (NSGSA).

This paper presents two approaches for multiobjective simultaneous coordinated tuning of damping controllers, a modified artificial immune network (MAINet) algorithm and a multiobjective immune algorithm (MOIA). The weighted-sum approach is used to handle the multiobjective optimization problem in the MAINet, while the Pareto-optimization approach is used in the MOIA. To investigate the ability of the proposed algorithms in designing the damping controllers, one small and one large power systems are considered. Two power-system stabilizers (PSSs) are designed for the small power system, while one PSS for a generator and one supplementary controller for a static var compensator (SVC) are designed for the large power system. The simulation studies show that the controllers designed by MOIA perform better than those by MAINet in damping the power-system low-frequency oscillations.