DETECTION OF STATIONARY FOREGROUND OBJECTS IN LOW-RESOLUTION VIDEOS USING DEEP LEARNING FOR ABANDONED OBJECT DETECTION

• Main Author: Saluky
• Format: Dissertations
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/84512
• Institution: Institut Teknologi Bandung

Monitoring using cameras that operate automatically to detect anomalies or significant events in the monitored area is a crucial element in security and surveillance systems. One significant challenge in these systems is the ability to detect and classify objects that have stopped moving, known as stationary foreground objects (SFO), which are often indicative of abandoned objects. Detecting SFOs is essential in various applications, including security, surveillance, and traffic management. The primary issue in detecting SFOs in low-resolution video is the degradation of video quality due to compression used to save storage and speed up data transmission. The low resolution of the video results in difficulties in identifying and classifying objects, especially when using conventional methods that require high-quality images for accurate analysis. Therefore, an innovative and efficient approach is needed to detect SFOs under low-resolution video conditions. This research aims to develop a method for detecting SFOs in low-resolution videos using deep learning technology. This approach offers significant potential in overcoming the limitations of conventional methods by leveraging the capabilities of neural networks to recognize complex patterns in low-quality visual data. The research results indicate that the proposed deep learning model can detect SFOs with high accuracy even in low-resolution videos. This model successfully identifies SFOs and distinguishes them from shadows or removed objects with minimal error. The study also compares the performance of the proposed model with several conventional methods and other existing deep learning models in the literature. The comparison results show that the proposed method has significant advantages in terms of accuracy and efficiency, particularly in low-resolution video conditions. The novelty of this research lies in the application of deep learning combined with tracking methods for SFO detection in low-resolution videos, which has not been extensively explored in previous studies. Additionally, this research proposes a model architecture tailored to the characteristics of low-resolution data, providing better results compared to previous models. The findings of this research contribute to the development of surveillance and security technology. With the ability to detect SFOs in low-resolution videos using tracking, surveillance systems can become more efficient and effective, especially in conditions where storage resources and bandwidth are limited. This study shows that the use of deep learning for SFO detection in low-resolution videos is a promising approach and can overcome various limitations faced by conventional methods. With the continuous advancement of deep learning technology, it is expected that this method can be widely adopted in surveillance systems, making a tangible contribution to improving security and operational efficiency.