ABSTRACT
Today the number of vehicles around is growing exponentially, which has been a cause for concern amongst the population living in major cities, metropolitan areas, and urban areas. As a result of the high traffic volume and limited space in dense urban areas, monitoring vehicles is particularly important [Batavia et al. 1997]. From real-time traffic management to supporting traffic planners, vehicle detection has many valuable applications. It is therefore difficult for law enforcement agencies to monitor every single vehicle. In addition to the rapid growth in the world’s population, the number of vehicles on the road is also rising every day. People, their families, and nations are adversely affected by road traffic accidents and injuries. In most countries, road traffic accidents result in a loss of 3% of the Gross Domestic Product (commonly known as GDP) caused by the death of approximately 1.3 million people each year [Sonka et al. 1993]. Despite the fact that there are several accident-avoidance methods available, deep learning-based accident prevention is gaining a lot of attention currently [Maity et al. 2022]. However, deep learningbased methods require more training images and also require more computational resources which may not be available in real-time. So, it is better to apply machine learning-based collision avoidance methods. The collision avoidance algorithm generally involves several steps, including detecting the vehicle and locating it using semantic segmentation, followed by recognizing the segmented objects [Nidhyananthan and Selva 2022]. Data from active sensors has been widely studied in the past few years for object detection. A more important consideration is the cost of these kinds of sensors. Our research focuses on passive sensors such as cameras to detect objects using vision-based methods. Images provide a more detailed picture of the environment than passive sensors. Object detection systems can be designed using the information conveyed by live images. Also, it is advantageous to have a higher resolution. Obtaining on-road information has never been easier, more affordable, and more accessible than with a camera. In addition to being able to use the data from these cameras to enhance intelligent security systems, these images can also be used to minimize traffic congestion through traffic management strategies [Bhattacharya et al. 2022]. Vehicle recognition (or classification) is very crucial for implementing intelligent transportation systems. The purpose of this technique is to identify moving vehicles and classify them accurately based on the flow rate [Arora and Kumar 2022].
