Developing Advanced Driver-Assistance Systems using computer vision and machine learning for driver safety use cases

Abstract

The automotive industry is increasingly using evolving technologies to increase driving comfort and safety. In addition to direct assistance systems for vehicle control (such as ABS, ESC, or ASR), there are also informational systems that warn the driver of the danger or provide the necessary information. Such systems include collision warnings, lane departure warnings, allocating pedestrians, and determining the distance to the car in front. These systems are considered Advanced Driver Assistance Systems and can be found as part of a high-end car kit or as an expensive option. Most of these systems depend on various devices and sensors, such as radars, LIDARS, and GPS modules. With the development of artificial intelligence and computer vision technologies, we have the opportunity to replace these components by creating software and obtaining only video data. This will significantly reduce the cost of production of such systems, make them more accessible for installation in cars that were not initially designed to include those functions, and will be able to supplement existing systems to increase accuracy and thus improve driving safety. This work aims to create a comprehensive system of driver assistance using only footage from the video of the road situation in front of the car and technology of artificial intelligence and computer vision.