oneAPI-Based Design and Development of an Advanced Driver Assistance and Monitoring System Utilizing Embedded Machine Vision Technology

Road accidents occur extensively, irrespective of the time of the day, with approximately over a million people losing their lives every year and thousands each day. Since the situations arise mostly from drivers being imprudent and/or getting distracted, a vision-based advanced driver assistance system is developed to warn them of any potential danger. The first proposed functionality can warn the driver of forward collision possibility with any obstacles ahead, including humans, animals, automotive, or non-motorized vehicles in proximity. If any vehicle tries to merge into the lane in which the vehicle equipped with the proposed system is moving, an appropriate warning is issued. In the second functionality, if the driver is using a mobile phone or drowsy, if they smoke or drink in the car, all are identified, and since these untoward actions might lead to distractions, the system issues a warning accordingly. The entire code is optimized employing oneAPI, an integrated software programming model offered by Intel, to enhance the performance of the system. oneAPI’s oneDNN library is specifically used to boost the system’s overall performance. The suggested method illustrates how oneAPI can speed up inference while producing accurate and transferable results on the affordable embedded Jetson Xavier NX platform. This proposed vision-based advanced driver assistance system can categorically aid to mitigate the accidents that occur daily in all places, especially India. The system is tested under all lighting conditions and is found to be perfectly stable.