A Deep Learning Based Approach to Enhance Object Edge Detection for Office Surveillance System
Keywords:Office Surveillance, Object Detection, Image Segmentation, Feature Extraction, Data Overlapping, Deep Learning, YOLOv8 Model
Office surveillance systems are essential for ensuring safety and monitoring activity in an office. These systems use a technique called object edge detection to track and trace objects during live action. Deep Learning is used in this study that improves object boundary detection and feature extraction in office monitoring systems. The most crucial aspect of the surveillance system is to detect and monitor human activity. As the advancement in video surveillance technology increases continuously, the fundamental task is to accurately detect objects by utilizing object detection techniques. Object detection techniques are widely used to achieve this goal by leveraging information gathered from images. By utilizing a stationary camera, the identification of humans in recorded videos can be done rapidly. This study uses the YOLOv8 model with two datasets, focusing mainly on the furniture dataset to detect table objects and changes that occur in them. Then compile and integrate datasets specific to office situations, including different illumination, object scale, and occlusion. Our proposed approach achieves excellent performance measures, including mAP50 with an accuracy of 99.3% on the Furniture Dataset. We comprehensively evaluate our approach to ensure its correctness. Also apply the YOLOV8 model to the Person Dataset to detect human activity without disclosing the individual’s identity, achieving an accuracy of 70.1%. To maintain a log of all object changes and authorized and unauthorized personnel, a database is created that contains all relevant information for security purposes.
How to Cite
This is an open Access Article published by Research Center of Computing & Biomedical Informatics (RCBI), Lahore, Pakistan under CCBY 4.0 International License