DEVELOPMENT OF MUSEUM VISITOR INTEREST DETECTION AND TRACKING SYSTEM BASED ON VIDEO PROCESSING IN LOW LIGHTING CONDITION

• Main Author: Reza Breitner, Alberth
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/86061
• Institution: Institut Teknologi Bandung

As a cultural institution, museums are responsible for protecting, developing, and utilizing their collections. To preserve their collections, museums rotate them based on visitor interest, with the expectation that more popular exhibits will attract a larger audience. Research on visitor interest is crucial to identify the most favored collections, yet methods such as surveys and direct observation often lack accuracy and can disrupt the visitor experience. With advancements in technology, visitor interest analysis can be conducted using sensors, Bluetooth signals, GPS, and edge computing. However, these methods require additional devices, which can limit coverage in large museum areas. Moreover, low lighting in museums often poses a challenge for accurately detecting visitor interest. This research aims to develop a visitor interest detection and tracking system with a system architecture based on video processing to overcome low lighting conditions. The system utilizes video from surveillance cameras as a data source, both live and recorded, with a focus on optimizing detection in low-light settings. To address the challenges of low lighting, the CLAHE video enhancement technique is employed. The system is built using a detection model developed from surveillance camera video data at the Museum of National Awakening. The area around the museum's collections (ROI) is used to measure visitor interest. During a one-week trial, the system successfully detected 1,221 interested visitors and 6,666 passersby. The test results show that the detection mode achieved an accuracy of 88.53% and an F1-score of 93.92%. The system also improved visitor interest detection by 25% using video quality enhancement. Usability testing indicated that 90% of 10 respondents were highly satisfied with the system. The findings of this study have the potential to be applied for more effective museum collection management, with recommendations for further development and testing in various rooms and other museums.