ADVANCED METHODS FOR MOVING OBJECT DETECTION IN PROBLEMATIC VIDEO FOR INTELLIGENT SURVEILLANCE SYSTEM

Abstract

In real-time environment, the detection of motion-based object is detected against challenging issues i.e. cluttered background. Practically, the desirable results are hard to obtain due to challenging issues such as dynamic background, illumination variation, floating or spouting water etc. The identification, detection, tracking and recognition is a crucial area of research in the area of computer vision that plays smart decision-based roles for intelligent video surveillance system, intelligent transportation, indoor-outdoor, virtual reality, medical diagnosis, robot vision navigation, law and enforcement, security, border monitoring, and many more. With the advent of technological advancements and availability of enormous data in the form of audios, videos, text and other forms. Artificial Intelligence emerged as one of the most significant technologies that simulates human intelligence through computers and several algorithms. Computer vision is one of the prime domains that enables to derive meaningful and crisp information from digital media. It includes several sub domains such as facial recognition, pattern detection, image classification, object detection and many others. Many vision-based applications such as traffic controlling, action recognition, human behavior analysis, industrial inspection, an intelligent surveillance are important issues for the researchers and industrialists in the modern days. This work presents the study of object detection in videos or continuous sequence of frames captured by the static camera. The thesis will make the clear image of comparison of object detection methods performed in prior researches. The main focus is on the real time application areas of video surveillance and the concerning challenges in the respective areas in near future. All the background modeling techniques and their sub categories studied and implemented by various authors is mentioned in literature. It also depicts the major challenging issues available in real-time environments. The significant aim of the proposed work is to develop an adaptive method to compute the threshold during run-time and update it adaptively for each pixel in testing phase. It classifies motion-oriented pixels from the scene for moving object using background subtraction and enhanced using post processing. The main objective is to focus and study the problematic video data captured through camera sensor, to handle challenging issues available in real-time video scene and to develop a background subtraction method and update the background model adaptively for moving object detection. Presently, the vehicle detection and tracking in intelligent transportation system is a highly active area. The proposed work also resolves the above problems and delivers solutions for the enhancement of the transportation system and the automobile industry. So, the work investigates a method using Google’s firebase platform (as a cloud service) and a background subtraction method for moving vehicle detection and tracking in a foggy environment. Here, the Google’s Firebox storage (Pyrebase API) is used for the storage of video that provides authentication and storage services. The moving vehicle is detected

using the background subtraction method for considered video followed by post- processing.

Detection and correctly tracking the moving objects in a video streaming are still a challenging problem. Due to the high density of vehicles, it is difficult to identify the correct objects on the roads. In this work, we have used a YOLO.v5 (You Only Look Once) algorithm to identify the different objects on road, such as trucks, cars, trams, and vans which results that the proposed approach attains improved results as compared to state-of-the-art approaches which is another contribution of the thesis.