ABSTRACT
The concern about road security and the protection of lives is increasing gradually due to intensified mobility, unfocused driving, and control over the vehicle. This interest led to development of advanced driver-assistance systems, which are intended to offer requisite information, caution, and instinctive intrusion to lower the risk/severity of an accident. The positional awareness of danger is the outcome of the automotive vision system. The input to the vision system is video streams collected from front, rear, and side cameras fitted on the vehicle. The quality of input received from the cameras is an important issue to be solved, but it remains a mostly unfocused area. Other challenges associated with the system are increased data rates, availability of memory, requirement of high-speed data processing and communication, and low-quality video streams. The low resolution of video streams is responsible for ineffective decisions made for the services provided by the system such as recognition of traffic light signaling, lane tracking, collision avoidance, and status of the driver.. The developments in the vision system and even machine learning-based automotive systems are mainly focused on services and not quality of input—i.e. video resolution. This study is a systematic look at the basics, recent developments, challenges, and scope of the vision system, with video resolution as the key parameter. New researchers in the area of image/video processing will benefit from the information provided and be motivated to contribute in this growing area.
