CROWD DETECTION USING CCTV MADE BY UNY STUDENTS

The trend of the COVID-19 pandemic in Indonesia has decreased. Nevertheless, the government strives to suppress this pandemic through various policies. Digitally, through the Ministry of Communication and Information (Kominfo), the government has used data on the movement of Mobile Subscriber Integrated Services Digital Network Number/MSISDN mobile phones from the Base Transceiver Station (BTS) to detect crowds of residents. Through this data, Kominfo can provide warnings via short messages in SMS blasts when there is a crowd of residents. Inspired by this, UNY students designed a crowd detection warning system based on the Deep Convolutional Neural Network using CCTV. They are Muhammad Nurwidya Ardiansyah (information technology), Muhammad Dzulfiqar Amien and Danang Wijaya (informatics engineering education) and Marifa Kurniasari (economic education).

According to Muhammad Nurwidya Ardiansyah, this system works to use CCTV devices as input media for video recording data in real-time, then detect people in the video frame. "After the object can be detected, the system will then define a crowd when there are two or more people with a distance of less than one meter," said Ardian. The calculation of the distance in the frame is carried out using the Euclidean Distance method. After the crowd is detected, the system will detect the color of the clothes of the people in the crowd so that the voice warning message issued by the speaker can be more specific.

Muhammad Dzulfiqar Amien said the crowd detection warning system based on a deep convolutional neural network is an innovation in technology development to suppress the spread of viruses made using three main components, namely the NVIDIA Jetson Nano microcontroller as a processing device, CCTV as an input device, and loudspeakers or speakers as an output device. The output of this prototype is in the form of a voice warning message to help remind the public to comply with health protocols, mainly to keep their distance and stay away from crowds.

Marifa Kurniasari stated that from the results of tests carried out on the prototype of this crowd detection warning system, the system could detect crowds at a speed of 22 frames per second. It could detect objects with an accuracy rate of more than 90%, and the crowd detection warning system was able to detect clothing colors so that the warning message given becomes more specific and increases the acceptance of the warning message. "In addition, this crowd detection warning system can also be run on 2 (two) CCTVs in real-time and simultaneously," said Marifa. (Dedu, Tj.Lak)

3. Good Health and Well Being
Category: 
Research
Indicator Name : 3.Good Health and Well Being: