Development of a Method of Terahertz Intelligent Video Surveillance Based on the Semantic Fusion of Terahertz and 3D Video Images

The terahertz video surveillance opens up new unique opportunities in the field of security in public places, as it allows to detect and thus to prevent usage of hidden weapons and other dangerous items. Although the first generation of terahertz video surveillance systems has already been created and is available on the security systems market, it has not yet found wide application. The main reason for this is in that the existing methods for analyzing terahertz images are not capable of providing hidden and fully-automatic recognition of weapons and other dangerous objects and can only be used under the control of a specially trained operator. As a result, the terahertz video surveillance appears to be more expensive and less efficient in comparison with the standard approach based on the organizing security perimeters and manual inspection of the visitors. In the paper, the problem of the development of a method of automatic analysis of the terahertz video images is considered. As a basis for this method, it is proposed to use the semantic fusion of video images obtained using different physical principles, the idea of which is in that the semantic content of one video image is used to control the processing and analysis of another video image. For example, the information about 3D coordinates of the body, arms, and legs of a person can be used for analysis and proper interpretation of color areas observed on a terahertz video image. Special means of the object-oriented logic programming are developed for the implementation of the semantic fusion of the video data, including special built-in classes of the Actor Prolog logic language for acquisition, processing, and analysis of video data in the visible, infrared, and terahertz ranges as well as 3D video data.