Real-Time Computational Efficiency Vehicle Detection and Counting Utilizing the Background Subtraction Technique and Non-Maximum Suppression Techniques

Abstract

By combining cloud computing, computer vision, and Internet of Things (IoT), it would be able to make the most of both sides. Because the IoT is mostly composed of connected, contained gadgets, it can store and process data gathered through the application of computer vision algorithms. It is able to achieve this by making use of the almost infinite resources provided by cloud organizations, including processing and storage services. The development and execution of a computer vision-based system are examined in this paper. that counts and identifies automobiles using machine learning (ML). The system consists of multiple stages, including initialization, background subtraction, object detection, bounding rectangles, vehicles counting and evaluation criteria. The proposed methodology first separates moving objects from the background and then employs a statistical technique called Mixture of Gaussians (MOG) for background subtraction to identify the automobiles in the image and Non-Maximum Suppression (NMS) to filter out overlapping bounding boxes to enhance the detection operation. The experiment's outcomes show how effectively cars can be found and counted. The result of the experiments using accuracy, precision, f1-score and recall are about 90% for the different types of video and from many corners.