DEVELOPMENT OF AN INTELLIGENT SURVEILLANCE SYSTEM FOR GESTURE RECOGNITION USING MACHINE LEARNING

Abstract

Over the years, video surveillance systems have been in use in various contexts such as traffic control, crowd control and protecting wildlife. In a dispensation characterised by security concerns and technological advancements, it has become necessary to develop intelligent systems for the purpose of surveillance. Intelligent video monitoring has become a vital tool for boosting security and safety in public areas. These systems combine the use of computer vision, machine learning, and artificial intelligence techniques to analyse video data and alert security personnel to potential threats. They can be configured to recognize particular behaviours, such as loitering, fighting, or unauthorized access to restricted areas, as well as to recognize particular objects, like persons or automobiles. The popularity of these systems has grown recently, owing to their ability to analyse video streams in real time and detect unusual conducts or events. However, traditional manual monitoring methods need a lot of manpower and are easily compromised. In addition, the cost of video surveillance goes up with mass data storage. This research proposes an Intelligent Surveillance System for Gesture Recognition. With sophisticated gesture recognition capabilities, this project focuses on enhancing security measures through cutting-edge technologies. The main objective is to identify potential threats in real-time, concentrating on spotting unusual conduct. The proposed system uses Media pipe to obtain data in real-time. Four different machine learning pipelines were trained for the gesture recognition. They include linear regression, Ridge Classifier, Random Forest Classifier, and Gradient Boosting Classifier, achieving an accuracy of 98.3%, 99.8%, 99%, 98.8%, a precision of 96.8%, 99.8%, 97,7%, 99.2%, a recall of 96.4%, 99.4%, 97.9%, 98%, and an F1-score of 96.5%, 99.6%, 97.8%, 98.6% respectively.