A Critical Study on Suspicious Object Detection with Images and Videos Using Machine Learning Techniques

Security has emerged as a significant concern in today's world, particularly in public places such as railway stations, airports, shopping malls, and crowded areas. Unattended object detection plays a crucial role in bolstering video surveillance systems. This paper presents an innovative approach to detect suspicious objects in images and real-time videos through the Internet of Things (IoT). Various object detection techniques are employed and compared for their performance. The study aims to enhance security measures by providing real-time monitoring and analysis. Various deep learning models, such as Faster-RCNN, Mask-RCNN, and Yolo, are compared for their effectiveness in detecting suspicious objects in images and real-time videos through the Internet of Things (IoT). The study highlights challenges like illumination changes, occlusion, noise, poor resolution, and real-time processing complexities. While machine learning methods show promising results, they still struggle with simultaneous recognition of multiple activities, leading to lower accuracy. The research also addresses current limitations, paving the way for improving the proposed surveillance solution.