Three-dimensional real-time detection of social distances by photogrammetric method with YOLO-V4 algorithm

Coronavirus (COVID-19) is a disease that first emerged in the last months of 2019 in Wuhan, the capital of China's Hubei province, and continues to be effective today. The disease spread in a short time due to airborne transmission and caused the disease to be declared as a pandemic by the World Health Organization (WHO) on March 11, 2020. The concept of social distance, which has come to the fore due to various epidemics throughout human history, has come to the fore again with COVID-19 and has taken its place in the first place in measures against the epidemic. In this process, various restrictions have been imposed for the control of social distance, control mechanisms have been developed and solutions integrated into auxiliary systems have come to the fore. Within the scope of the study, an algorithm was designed for real-time detection of social distances with three-dimensional coordinates using photogrammetric methods. The scenes of the images obtained with the stereo vision system created by two computer cameras were evaluated and the people were identified with the YOLO-V4 algorithm. Social distances between detected persons were calculated on the basis of real three-dimensional coordinates and analyzed in real time. When existing algorithms for social distance detection were examined, it was seen that distances were calculated based on two-dimensional pixel coordinates or scene-dependent three-dimensional reference points. The developed algorithm calculates interpersonal distances with high accuracy, independent of the scene, with three-dimensional real coordinates, and in this respect, it presents a different perspective from other studies in the field.