The present invention relates to a method and a system for recognition of objects near a ship by using a deep neural network to prevent a collision with the object by recognizing a neighboring object that may be risky to the ship sailing in a restricted condition such as a foggy environment. All object movements within a predetermined radius are detected and recognized so that collision accidents with objects on the sea in an environment such as fog caused by bad weather at sea can be prevented, and a risk alarm is notified to a captain when the object is detected so that collision accidents can be remarkably reduced. In addition, peripheral environments are detected by only installing a CCTV camera so that expenses can be reduced, human negligence can be prevented, and the system can be easily constructed to prevent collisions.