ENHANCED SURVEILLANCE SYSTEM: MASKED FACE RE-IDENTFICATION USING CLOUD COMPUTING AND DEEP LEARNING

Abstract

The spread of Covid-19 virus, and the growth of suspicious people existence, raised the difficulty of securing public enterprises and pivotal organizations. In order to limit infection and prevent intruders from entering, the availability of qualified monitoring systems has become necessary. This paper proposes a surveillance system based on Cloud computing that observes people entering public buildings. The system's goal is to employ deep learning to reveal masked passers who have been infected with viruses or recorded as invaders. It considers dealing with images of different accuracy. It focuses on excluding important details from the exposed part since a large amount of information was lost due to the covering part. The system has been tested first offline using MATLAB 2020, then it was implemented online using Python. Both versions use Resnet CNN. This resulted in similarity rates while identifying banned ranges from 75% to 100%. Five computed performance measures reach 100% for men and 99.5% for women, when training to validate percentage was 3:1, except for one person due to image dispersity. However, when the ratio is 2:3, the average score, excluding those of similar people, is around 97%.