Reliability of camera systems to recognize facial features for access to specialized production areas

摘要

The article deals with ergonomics and reliability of camera systems for recognition of facial features and identify person for access to specialized areas. The monitoring of areas relates not only to crime, but it is also an integral part of access to specialized production areas (pharmaceutical production, chemical production, specialized food production, etc.). It is therefore important to adequately secure these premises using the relevant system. One of them is a system based on user identification using specific facial features. For this purpose, there are CCTV systems for recognition of facial features of different price categories (conventional cameras, semi-professional and professional) on the world market. However, problematic situations may occur when identifying. For example, by having the user partially masked face. This research is focusing on the problem. The main goal of the research is establishing the scale of negative impact, in case the identified person has partially masked face, on camera systems recognizing facial features, primarily on recognition time. The results are evaluated in detail. Some camera systems are not suitable in specialized production areas due to their insufficient recognition ability. From all the tested devices, the HIKVISION iDS-2CD8426G0 / F-I camera identification system has proved to be optimal for identification purposes. In the case of designing, it is therefore necessary to choose suitable camera systems that have ergonomics and reliability at a level that will guarantee their sufficient use in the mentioned areas, while decreasing comfort and user-friendliness as little as possible. By measuring the ergonomics and reliability of these CCTV systems, it can be stated that there are statistically significant differences between conventional, semi-professional and professional systems, and it’s not just a design change, but also a more efficient recognition method.