Semantic Analysis
This chapter describes the second and most important original contribution of this book namely a methodology to implement systems suitable for high semantic level analysis of video streaming recorded into a narrow domain. This methodology is independent b
- PDF / 957,333 Bytes
- 21 Pages / 439.37 x 666.142 pts Page_size
- 68 Downloads / 188 Views
Semantic Analysis
This chapter describes the second and most important original contribution of this book namely a methodology to implement systems suitable for high semantic level analysis of video streaming recorded into a narrow domain. This methodology is independent by the method proposed into the previous chapter. Since the latter improves the performance of the object tracking algorithm, the former works with or without the latter. This methodology can work also using single camera systems. The chapter starts with a brief problem overview and then presents the proposed methodology.
5.1 Introduction As shown in Fig. 3.1, the semantic analysis of human behavior is the last stage of a complex processing chain. This chapter presents the grammar based methodology proposed in this book. This methodology allows for a hierarchical analysis of the recorded scene. According to the level of details used in scene recording, this methodology can provide from a semantic analysis of the whole scene till a detailed behavior analysis of a single person. From this perspective, this methodology can be seen as a unifying approach to the three classes of methods described in the Chap. 3. The grammar based approaches, as shown in Chap. 3, allow for human behavior recognition and classification. The ability to define dynamically, for example using some clustering methods, the classes of the observed scenes is not present in other powerful tools used in literature such as the Hidden Markov models (HMM). The proposed methodology starts from the idea at the base of the scene interpretation systems that try to interpret the scene studying the trajectories of some relevant points of the moving objects (their barycenters). The idea behind this kind
A. Amato et al., Semantic Analysis and Understanding of Human Behavior in Video Streaming, DOI: 10.1007/978-1-4614-5486-1_5, Springer Science+Business Media New York 2013
57
58
5 Semantic Analysis
of approaches is the mapping between trajectories and behaviors. The ratio can be synthesized into the observation that in order to accomplish a given task one must follow a prefixed series of movements. In this book, the trajectories are represented by means of string of symbols (the alphabet of the domain specific grammar) each one having a semantic value into the specific domain. The mapping among symbols and the semantic meaning of areas of the scene is done manually at design time. For each domain of application, a grammar is defined in order to specify the recognizable sentences of that domain. In this way it is possible to define a correspondence between the set of the sentences writable with this grammar and the set of allowed actions in the scene. Since the mapping among portions of scene and symbols/semantic meanings and the grammar are specific for each application, this system belongs to the class of narrow domain systems. In general, the main drawback of the existing approaches to semantic analysis of the human behavior, even in narrow domains, is inefficiency due to the high
Data Loading...
