SAFE AND SECURE PLATFORM DEVELOPMENT FOR CLIMBING RECOGNITION AT KAI STATION

• Main Author: Kevin Cahyadi Giri, Nyoman
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/66348
• Institution: Institut Teknologi Bandung

Railway station development must move towards smart station to support smart city development in Indonesia. Safe and secure aspect needs to take into account when developing smart station. One example of an activity that threatens the safe and secure aspect is climbing fence. Climbing fence can damage the fence itself or lead to violence. Climbing fence can be suppresed with video analytics. Video that will be analysed comes from installed CCTV at the railway station. Video analytics will run on VIANA server with OpenPose and Deep Neural Network algorithm to recognise climbing fence action in railway station. The test on OpenPose and DNN is conducted in production environment which is in Bandung railway station. The outputs of the test are accuracy, precision, and recall. The results of the test are OpenPose and DNN has 75% accuracy, 76% precision, and 72% recall. This algorithm also use resources without sudden significant increment. This algorithm use 35% GPU, 8% memory, and the temperature of GPU is 63o C. The prediction result of this algorithm is shown in dashboard. The dashboard will show the map in railway station and also area that captured by CCTV. For every area, dashboard will show pictures that have been indicated has climbing fence activity captured on them by the algorithm. The pictures are stored on storage in VIANA.