Method and apparatus for predicting a vehicle collision with a surrounding object

A method of predicting a collision between a vehicle with a surrounding object moving around the vehicle while traveling, the method comprising:

• obtaining, by a surrounding object information detection unit, current state information on the object through a sensor equipped on the vehicle and obtaining, by a lane detection unit, information on a lane on which the vehicle travels;
• calculating, by a risk level determination unit, a collision risk level between the object and the vehicle by using state information on the object;
• predicting, by a surrounding environment prediction unit, future state information on the object by using an estimation algorithm or a deep learning algorithm;
• generating, by an information integration unit, a simplified bird's eye view by integrating the current state information, the lane information, the collision risk level, and the future state information;
• predicting collision modes through a convolutional neural network (CNN) using the simplified bird's eye view as an input;
• and
• calculating a probability of a result of classifying the collision modes and outputting a final collision mode determination result when the calculated probability exceeds a threshold value,
• wherein the predicting of the collision modes includes
• extracting a collision prediction feature through a convolutional layer and a pooling layer, and
• classifying the collision modes through a fully connected layer.