Real-time Crowd Detection of Vehicles: Approaches and Open Challenges

Real-time vehicle crowd detection has become an urgent necessity technique for many applications like traffic control management, monitoring, and self-driving. Traffic crowds sometimes happen because many vehicles are located in the same place or take the same route without alternative routes. This leads to unwanted crowding, even though alternative paths can be taken. Many recent technologies have been suggested to solve crowd issues such as smart traffic lights and optimized GPS routes and others. However, in-vehicle crowding detection solutions are rarely based on embedded devices using machine learning algorithms. Smart embedded engineering systems with limited resources can challenge implementing efficient crowded vehicle detection algorithms. Recent machine learning technologies have tried to overcome these challenges. This review paper aims to provide a clear overview on embedded systems, applications, case studies, used datasets, online development platforms, machine learning approaches, and available frameworks for Lite deep and machine learning algorithms to be a valuable resource for researchers and practitioners alike allowing to suggest both create image dataset and smart embedded system of crowd vehicle detection. A list of an existing vehicle crowd detection approaches systems and technologies has been listed in a table for comparison. Various methodologies, open challenges, and future directions in these domains are also discussed to be utilized in future works.