SAFE AND SECURE PLATFORM DEVELOPMENT FOR THROWING RECOGNITION AT KAI STATION

• Main Author: Ian Widi Nugroho, Vincentius
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/66356
• Institution: Institut Teknologi Bandung

Safe and secure is one of the most important aspects in the development of smart cities but is often overlooked. Smart mobility like on KAI train stations is also one of the part of smart cities therefore its safety and security aspects of its facilities needs to be put on attention. Attempts on increasing safety and security in KAI stations could be done by done by detecting activities which could potentially harm safety and security automatically by using CCTV and video analytics. Right now, ITB, KAI, and LPDP have developed a video analytics platform called VIANA. However, there are other safe and secure threatening aspects that haven’t been handled by VIANA. The objective of this final projects is to answer the problem how to develop VIANA by adding new necessary safe and secure features such as fence climbing, trespassing, and item throwing. The output of these features will be a safe and secure index that is used to determine the level of safe and secure of one area. This final project will only focus on the addition of throwing detection feature. Throwing is one example of dangerous activity in train platforms. This final project uses Design for Six Sigma methodology by Kosky. Its flow started with problem definition that is the existence of people throwing stuffs in train platforms that could harm other person. Next, exploration is done to find the functional and nonfunctional requirements for throwing detection based on processing speed and GPU performance. OpenPose algorithm that can detect actions based on the position of human body parts is chosen to be the best way to detect actions. A model from OpenPose has also trained which then become the base of inferencing that is done by the throwing detection feature. A dashboard is also developed to show the results of throwing, climbing, and trespassing detection features. Throwing detection feature has successfully been added to VIANA and used to detect throwing scenarios in KAI station platforms. Based on the feature’s testing, the feature could be run for a long time nonstop with high precision, recall, and accuracy. This proves that throwing detection feature could be used by KAI. Even so, throwing detection feature could still be developed by adding datasets of photos for other actions such as calling on the phone and also by adding more photos of throwing from various angles. Another aspect that needs to be kept in mind for the fluency of detection is having stable network in getting data stream from CCTV.