Intelligent Transportation System for Multiple Vehicle Detection and Tracking Using Multi Variant Feature and Machine Learning Techniques

Abstract

High traffic loads have significantly increased due to the population growth, and street newlinetraffic is considered as a major problem. As a result of the large number of vehicles that newlinemove daily from one location to another in big cities is a miserable but preventable newlinecompanion. In various real-life performances, the vehicle and event identification in newlineaerial video sequences plays a crucial role particularly, in cases of traffic collision and newlinecongestion. Yet, it can be difficult to assess aerial data from the video connected to aerial newlinevehicles and properly classify aerial vehicle data. The goal of this research is to identify newlinewhich classes of aerial video frames each class belongs to. To achieve the highest training newlineefficiency, multiple layers, and building elements are used to create the CNN architecture. newlineThe Selection and the Decision Network (SeDeNet) and the Classification Network newline(ClsNet) model are two separate sub-models of the CNN architecture, which is used to newlineextract the most accurate information from the presumptive image data and generates newlinefeature weights that are fed into the ClsNet model to validate high-performance training newlineand maximize classification outcomes. The CNN-based SeDeCls proposed model is newlineverified based on the Video Dataset such as Event Recognition Aerial (ERA) to calculate newlinethe consequences of performance. The model accurately detects which particular aerial newlinevideo scene frame belongs to which class are used as evaluation metrics for the newlineperformance comparison against varied classification models such as accuracy, precision, newlinerecall, and F1 score. Thus, the developed CNN-based SeDeCls model shows greater newlineperformance than varied typical classification models. newline