A Systematic Review on Vision-Based Traffic Density Estimation for Intelligent Transportation Systems

Traffic congestion is often considered one of the major challenges faced in urban areas. It is important to address this issue due to its significant negative impacts on both society and the environment, including decreased productivity and increased pollution. For this reason, implementing a traffic density estimation system is necessary as it can be further integrated into adaptive traffic control systems that dynamically adjust traffic lights based on real-time congestion levels. Different from existing papers that categorise vision-based traffic density estimation methods into microscopic and macroscopic approaches, this paper contributes a novel taxonomy by introducing hybrid approach, which combines the two to leverage their respective advantages. Furthermore, this review paper offers guidance for future research on this topic. Later in the discussion, the three approaches for estimating traffic density will be broken down into specific methods used, namely image processing techniques, machine learning models, deep learning models, or a combination of them. This paper also provides a coherent discussion of the details of these papers, as well as their advantages and drawbacks. To the best of our knowledge, this is the first review paper that specifically discusses traffic density estimation methods based exclusively on image and video data.