Vision-Based Data-Driven Modeling Vehicle Detection in Videos Using Convolutional Neural Network

Vision-Based Data-Driven Modeling Vehicle Detection in Videos Using Convolutional Neural Network

R. Regin, Sriraam Ramesh, Athiyan Ramesh Kumar, Praghalad Krishna Gandhi, Rubin Bose S
Copyright: © 2023 |Pages: 20
DOI: 10.4018/979-8-3693-1301-5.ch011
OnDemand:
(Individual Chapters)
Available
$37.50
Current Special Offers
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

Object detection is a vital component for autonomous driving, and autonomous cars rely on perception of their surroundings to ensure safe and robust driving performance. It shows how the perception system makes use of object identification algorithms to precisely identify nearby items like pedestrians, cars, traffic signs, and barriers. It goes on to say that detecting and localising these things in real-time depends greatly on deep learning-based object detectors. The most recent object detectors and unresolved issues with their integration into autonomous vehicles are also covered in the essay. It mentions that deep learning visual classification methods have achieved enormous accuracy in classifying visual scenes; it makes use of the convolutional neural network. However, it points out that the visual classifiers face difficulties examining the scenes in dark visible areas, especially during the nighttime, and in identifying the contexts of the scenes.
Chapter Preview
Top

1. Introduction

Modern civilization now considers traffic monitoring to be essential, and using cameras and computers to accomplish this aim is becoming more and more common. However, in the modern era of advanced technology, manual video analysis is not only time-consuming but also impractical (Khalifa, et al., 2014). In order to create intelligent traffic monitoring and management systems that conserve resources and need the least amount of human interaction, computer-based image processing technology has been created (Mohamed, & Mesbah, 2016). These systems can provide sophisticated and workable monitoring programmes by analysing video pictures, detecting and identifying vehicles, and taking necessary action in the event of unexpected circumstances (Sadek, et al., 2021).

The supervisory department now uses computer connection monitoring methods to accomplish the aim of intelligent traffic monitoring and management (Abbassy, et al., 2020). These systems analyse video from cameras, identify cars, and do standard management activities automatically using computer and image processing technologies. They can also spot anomalous circumstances and respond to them in the best, most secure way possible (Derindere Köseoğlu, et al., 2022). This approach is appealing to academic and corporate circles throughout the world since it saves a sizable amount of labour and material resources. The vehicle that uses vision. The main approaches for object detection include classic machine vision techniques and sophisticated deep learning techniques (Ead & Abbassy, 2018). The former, which may be categorised into three different approaches the technique of backdrop subtraction, the method of continuous video frame difference, and the method of optical flow uses a vehicle's motion to distinguish it from a stationary background picture (Jain, et al., 2022). The moving foreground region is divided by the threshold using the video frame difference approach, which computes the variance based on the pixel values of two or three successive video frames (Arslan, et al., 2021). This technique may also detect the car halting by muzzling sounds (Bhoumik, et al., 2020).

Complete Chapter List

Search this Book:
Reset