Ship Ranging Based on Monocular Imaging and Applications

Environmental perception, including object recognition and distance estimation, is a crucial aspect in the field of computer vision. Although distance estimation has received relatively less attention compared to object recognition, its accuracy becomes particularly crucial in specific scenarios, such as when ships are entering ports or gates. In such situations, precise distance control is essential, as errors could lead to the risk of collisions between the ship and the dock. Due to significant variations in the sizes of ships, traditional distance estimation methods often exhibit substantial errors. To mitigate these inaccuracies, this paper proposes a self-adaptive ship size estimation approach. In addition to introducing an adaptive model, we further design a distance estimation model incorporating tilt information to enhance the accuracy of distance estimation, especially for objects closer to the camera. Experimental results demonstrate the effectiveness of this approach in improving the accuracy of distance measurements, particularly in overhead perspectives.