A Vision-based Ship Speed Measurement Method Using Deep Learning

In the field of water transport, ship speed estimation has become an important topic in the research of intelligent shipping systems. Traditionally, Automatic Identification System (AIS) is used to extract the ship speed information, however, busy canals will lead to the loss of AIS data information, especially some small ships that are not installed with AIS systems, which will lead to various accidents. To compensate for this shortcoming, we propose a method to extract ship speed using vision-based deep learning techniques. Firstly, We learn to recognize ship models by training the YOLOv7 convolutional neural network, and the DeepSORT algorithm is used to combine the ship motion features and depth features to track the ship targets in the video in real time. Secondly, an adaptive mapping model from pixel space to the real world based on semantic segmentation is designed to solve the uncertainty of UAV flight height and shooting angle. Finally, based on this mapping, we use the change of ship motion pixels to calculate the ship’s speed and count the ship traffic. To improve the accuracy of the speed estimation we use the least squares method to fit the data measurement. Exhaustive experiments were carried out in various scenarios. Results verify that the speed measurement accuracy of the proposed framework is above 93% in different scenarios, which has excellent performance. This paper also paves a way to further predict ship traffic flow and ship trajectory in water transportation.