Article Preview
TopIntroduction
Computer vision (CV) is a field that features the strategies of acquiring, processing, analyzing, and understanding the images in general, of high-dimensional data from the real world data (Esteva et al, 2021; Ahmed et al, 2020; Hosny et al, 2020). The major attention on using computer vision is to produce or to generate more representative information from the high dimensional data. The field of CV has been developed by imitating the skills of human vision by electronically perceiving and understanding a picture. The understanding of image has been made with the help of Mathematics, Physics, Statistics and Learning theory (Forsyth, and Ponce, 2011; Fouad et al, 2021; Hegazi et al, 2020; el den Mohamed et al, 2020). The images can be acquired from several formats such as video sequences, views from multiple cameras (3D images), multidimensional data from a medical scanner, etc. This evolving technique has been constructed as a system and used in various domains of applications such as autonomous cars, bikes or in simple terms autonomous vehicles (Kisačanin, 2017; Mahmoud et al, 2020), human activity recognition (Plötz and Guan, 2018; Ramanujam et al, 2021), autism children (Rudovic et al, 2018) etc,. In specific, the development of computer vision has shown abrupt and appreciable improvement in remote surveillance systems. This trending remote access on surveillance and intelligent analysis of computer vision techniques has replaced the traditional security personnel.
Remote surveillance usually involves live monitoring of sensitive areas on installing closed circuit television (CCTV) cameras (Ratcliffe et al, 2009). In recent days, various types of cameras have evolved such as Day/night or Infrared cameras with varifocal lens, Network/ IP cameras, wireless, PTZ/ Speed Dome cameras, High-definition (HD) cameras, etc. This type of camera helps to detect the criminals and suspects with evidence and they are also used for recording the movement of people and property.
Nowadays, the large network of cameras is predominantly installed in public places like Airports, Railway stations, Bus stands and Office buildings (Caplan et al, 2011). These networks of cameras provide large video data which is monitored manually and may be utilized only after the fact. Studies have shown that a person assigned to sit in front of a video monitor for several hours a day and watch for particular events is an ineffective security system. Tests have demonstrated that after only 20 minutes of watching and evaluating monitor screens, the attention of most individuals has degenerated to well below acceptable levels. Monitoring video screens is both boring and mesmerizing. This limits the ability to distribute information across an enterprise/ businesses/ healthcare, which could help to minimize company-wide threats and alerts in real case scenarios.
Fascinatingly, automated analysis of such huge video data can improve the quality of surveillance by processing the video faster. Such automated analysis is more useful for high-level surveillance tasks like suspicious activity detection or undesirable event prediction for timely alerts or behaviors or events of interest.
For example, losing their beloved child is the worst nightmare for every parent. Sometimes, within a fraction of a few seconds their child may go somewhere else in the crowded area. If it happens at home or in a less crowded public place, it will probably be easy to identify the child in some corners or immediate vicinities (Faruqui and Yousuf, 2020; Yusuf, 2020). But in medium/ highly crowded public areas such as temples, streets, malls it is a challenging task. There are so many fascinating reasons that may distract the children and to get lost from the control of the parent (shin et al, 2014). Thus, this paper addresses the global wide challenge on detecting a missing child from the crowd.