Article Preview
TopIntroduction
Computer vision algorithms have played a vital role in video surveillance systems to detect surveillance events for public safety and security. Even so, a common demerit among these systems is their unfitness to handle diverse crowded scenes. In this paper, we have developed adaptive crowd behavior and motion detection algorithms using fuzzy logic. these solutions deal with some of the problems encountered in smart video games (Fradi, 2017; Chen, 2015; Li, 2015; Burghouts, 2011; Ullah, 2013; Wang, 2016).
Intelligent video surveillance is a branch of computer vision. Expresses is itself a broad research axis, applied in different fields. Much research is already being done in this area. In particular, the recognition of activities and behavior in a video are subjects currently investigated by several researchers (Ko, 2008; Chebi, 2015; Chebi, 2016). In crowd scenes, three types of problems are commonly posed: (i) motion pattern extraction; (ii) event detection; And (iii) estimating flows. These problems are not new and have been addressed in several studies (Robert, 2000; Baumann, 2008; Morris, 2008; Chebi, 2015) (Chebi, 2016). Through this state of the art, we describe the descriptors or types of information exploited to deal with each of the three problems in order to arrive at a set of information characterizing these problems in a common way.
The approach suggested in this article given in Figure 1 differs from the existing approach (Ullah, 2013; Wang, 2016; Ko, 2008) by its dynamic of detecting anomalies in which it makes possible the detection of anomalies for both cases (the case of a group or a single person).
The approach total used in this research task for the detection of anomalies is characterized by its dynamic mechanism making it possible to detect in an automatic way the processes of anomalies “case of a normal and abnormal behavior.” It can be divided into six stages to gather into three sublevels (Figure 1): the bottom level which estimates the optical flow, the intermediate level which constructs of the model magnitude and orientation and uses the techniques of image processing, and the semantic level which notifies of the operators.
Our work in this article deals with problem relates to the analysis of crowd behavior. We describe the proposals brought in the way following: