Automatic recognition of dangerous objects in front of the vehicle’s windshield

In this article, an automatic recognition system, of objects that threaten the safety of vehicle users is proposed. These objects are detected in front of the vehicle’s windshield based on images from video recorders and then classified. The tests were conducted using artificial neural networks on a specially prepared database of frames from real recordings. Recognized objects include: car wheels, car parts, birds, stones and ice. They were recognized in single- and multi-class detections for various parameters of the network model. Detection efficiency of $95 \%$ was achieved in the best case.