Blind-Spot Monitoring Using Deep Learning.

PONNARASAN, T.; KUMAR, T. R. RANJITH; RUBAN, I.; SUMITHRA, M. G.

In recent years, the rapidly increasing vehicular population, driving environment and human factors has led to a lot of traffic accidents. If there exists a mechanism to detect obstructions in the road, and then relay the processed information back to the driver. He may be alerted about the impending danger. This paper proposes a vision-based monitoring algorithm to detect vehicles in a blind-spot area using three rear-view cameras. Three frames are stitched to obtain a single wide-range frame. The stitched frame eliminates a major portion of the vehicle's blind-spot. This system uses the YOLO algorithm to detect a vehicle in the detection window. The alarm signal is generated if the detected vehicle in the blind-spot crosses the hazardous zone. Vehicle data set with truck, bus, car, and motor-cycle is used to train the YOLO model. The processing speed of the system is improved while keeping performance degradation as minimal as possible.

TRAFFIC accidents; TRAFFIC monitoring; AUTOMOBILE rear view cameras; DEEP learning; MOTORCYCLES

International Journal of Pharmaceutical Research (09752366), 2020, Vol 12, Issue 1, p1403

0975-2366

Academic Journal

10.31838/ijpr/2020.12.01.232