CCTV CAMERA-BASED TRACKING USING DATA SCIENCE TECHNIQUE: A DEEP LEARNING APPROACH

Abstract

This survey is a compilation of results from 20 research papers related to smart surveillance systems based on the integration of IoT, computer vision, deep learning, and embedded technologies. Traditional CCTVs are constrained due to manual monitoring and lack of intelligent response. In contrast, modern frameworks are based on the usage of Raspberry Pi [1][2][3], PIR sensors [4][5], Pi cameras [6], wireless IP networks [7][8] for real-time motion detection, facial recognition [9][10][11], intrusion alerts [12][13], remote access. Techniques like CNN [14], LBPH [15], Haar cascade classifiers [16], SSIM [17] and NSDT [18] help in increasing automation and accuracy. Applications cover areas of smart homes, traffic monitoring [19], gender sensitive urban safety [12], and crowd behavior analysis [20]. These systems exhibit a range of features that are applicable to a system that can be scaled up and down, is economical, and can be adapted to work in many different environments. Collectively, the reviewed literature reveals a change in paradigm toward proactive, intelligent surveillance ecosystems requiring the least amount of human intervention and resulting in maximum situational awareness, public safety, and operational resiliency.