Study of a Method for Effective Noise Suppression in Passive Personnel Screening Systems

The paper discusses approaches to increase the sensitivity of passive personnel screening systems by integrating sequential frames with a moving subject. Several state-of-the-art methods of computer vision are considered for this purpose, which can be used to track moving subjects even on very different frames. The results of experiments using the computer vision method DensePose, based on the use of artificial neural networks, are presented. Using DensePose, the segmentation of a moving subject and textural UV-coordinates for the surface model of the human body are found on frames, which are used in the described frame-by-frame integration method. Considering obtained results, the shortcomings of the proposed frame integration method are identified and listed. The directions of further research are suggested.