Penerapan Mediapipe Pose untuk Pengenalan Isyarat Tangan dalam Situasi Darurat di Ruang Publik

Abstract

CCTV is a location monitoring information system that uses video cameras placed in a specific location and connected to a closed network so that the control room can be monitored. The need for CCTV continues to grow along with the increase in crimes that occur in public areas. In 1968 in the city of New York, United States CCTV began to be installed on busy main streets to minimize crime. CCTV began to be commonly used as a surveillance solution and reduce crime in public places. In this research CCTV monitoring is made easier because officers who monitor CCTV do not have to be in the control room to find out if there are people asking for help, especially in quiet public areas and the reactions of officers are faster in providing help. CCTV utilizes computer vision technology to recognize hand signals that are used as a sign of someone asking for help. The hand gesture used is to raise both hands in the air because this gesture is commonly used to get people's attention. This study utilizes the MediaPipe framework to recognize a person's hand signals by detecting and providing coordinate points (landmarks) on a person's body then the coordinates are processed so that they match the specified hand signals. This research was tested at outdoor locations with time parameters, namely morning, afternoon and evening and with distance parameters for each time, namely 3 to 6 meters. The results of this research obtained the best value at a distance of 4 meters during the day as evidenced by an accuracy value of 100% and an F1-score of 100%.