Research And Application Of Ship Passage Monitoring And Measurement In Inland Waterway

On the inland waterways of our country,there have always been shortcomings in the monitoring and measurement of ship traffic.The reason is that our country has a vast territory,numerous waterways,and a large number of ships,which makes management difficult.At the same time,the ships on the waterway did not fully use AIS,which caused the waterway department to be unable to conduct monitoring and measurement,and could not effectively count the ship data.This series of phenomena reflected the current problems faced by the waterway department.At first,people used traditional motion detection algorithms such as background subtraction,difference method,Gaussian model method,etc.,combined with contour detection algorithms to find moving ships,and then monitor the ship’s dynamics.However,traditional algorithms are susceptible to interference from the surrounding environment,such as light,weather,etc.These factors make the algorithm unable to run stably and continuously.With the advent of artificial intelligence,convolutional neural networks have begun to be applied on a large scale in image and video.Among them,convolutional neural networks represented by the YOLO series have the advantages of fast speed and high accuracy,and are the first to be applied in the industrial field.Convolutional neural networks do not need to specifically design feature extractors,but through end-to-end training and learning to extract abstract features of the image,so that it can recognize and locate objects in the image.At the same time,in the face of complex and changing backgrounds,the network has strong robustness.Based on the YOLOv4 Tiny network,this paper improves the multi-target recognition network into a single-target recognition network,and by loading pre-training weights,it merges real ship image data for migration learning,which improves the accuracy of network recognition.And by improving the recognition method and dynamically adjusting the recognition frequency,the work efficiency of the network is greatly improved.In terms of counting,the IOU tracking algorithm is used to track the target,and the traditional adjacent frame coordinate subtraction counting is abandoned.The rapid rejection and straddle test algorithm can be used to more quickly and accurately determine whether the target crosses the boundary line.At the end of this article,a novel method of measuring the deadweight of a ship is proposed,which can quickly estimate the deadweight of a ship.