SUBSYSTEM POSE ESTIMATION AND OBJECT DETECTION WITH GRAPHICAL USER INTERFACE ON BEHAVIOR DETECTION FOR SECURITY SYSTEM BASED ON SECURITY LEVEL IDENTIFICATION

• Main Author: Chandra, William
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/41668
• Institution: Institut Teknologi Bandung

The security system in the house currently is being monitored manually. This is quite burdensome for homeowners to be able to maintain the condition of the house in a safe condition at all times. Hiring security to do 24-hour surveillance will cost a lot. In this final project, created a security system that can automatically identify the level of home security automatically based on behavior analysis. Because the advance of image identification technique become more accurate and fast. This system is made using 4 IP cameras and 1 computer unit. The behavior analysis system is implemented using the pose estimation with mobilenet with accuracy 81.75% accuracy when camera implemented at top 77.54% and can work on speed 18 FPS, object detection with YOLO with accuracy 64.6% along with a graphical user interface using windows form with photo and recording feature. Component in this system will be mixed to create the behavior analysis. The results of the test show that behavior analysis algorithms have been made have enough speed and accuracy in the implementation of security systems based on behavior analysis.