Design And Implementation Of Ship Identification System Based On YOLOv5 And Beidou Short Message

Guangxi,as a coastal province in the south of China,is well developed in water transport.There are a large number of cruise ships and fishing boats in inland waterways,and there are also rich types of maritime ships on the sea surface such as Beibu Gulf.In order to ensure the safety of ship transportation,the administrative department usually sets up some high-altitude lookout type high-definition surveillance cameras on the banks of inland rivers,coasts and islands to monitor the operation of ships in real time,discover potential safety risks in time,and alert the administrative department and nearby ships through early warning.However,the above video surveillance equipment is not highly intelligent and is usually installed on remote coasts,banks and islands.Due to the lack of4 G and other conventional communication signals,it is easy to make early warning communication difficult and the cost of rebuilding communication facilities is high.Beidou short message communication belongs to satellite communication,which uses Beidou short message to transmit early warning information.Although it is not restricted by traditional communication facilities,the single communication length of Beidou communication has certain limitations,which makes it difficult to transmit ship pictures.Using traditional image compression processing,it is impossible to transmit pictures taken by high-altitude observation surveillance cameras through one Beidou short message.Aiming at the problem of intelligent identification and transmission of early warning information of high-altitude lookout-type HD surveillance camera equipment targets in remote coastal,riverbank and island areas,this thesis carries out research on deep learning-based ship target identification and Beidou short message based early warning information transmission technology,respectively,and the research contents and results of this paper are as follows:(1)Aiming at the problems of low intelligence in recognizing ships by video surveillance equipment and slow inference speed of target detection network,a lightweight ship target detection network based on YOLOv5 model improvement is proposed.Firstly,we use Labelimg software to create ship dataset under surveillance images and enhance the dataset,and then use Ghostmix network module,depth separable convolution,and fast pyramid pooling module to build a lightweight target detection network,and also use cooperative attention mechanism to improve the accuracy of ship detection.The experimental results show that the accuracy,recall,F1 score,m AP,and FPS of the proposed network reach 94.7%,92.2%,93.4,95.2%,and 8 frames/s,respectively,which are 10.7%,10.5%,10.60%,and 13.7% higher than the base YOLOv5 network in terms of accuracy,recall,F1 score,and m AP metrics,respectively,and inference speed is improved by 17 frames per second.(2)In response to the limitation of the single communication length of Beidou short messages and the difficulty of transmitting suspicious ship images,YOLOv5 s is proposed to identify ship images and extract summary information.The Beidou communication transmission protocol is studied,and a Beidou short message merging mechanism and timeout retransmission mechanism are designed.Text summary information is encapsulated according to the Beidou communication transmission protocol,and the feasibility of early warning transmission using the Beidou communication transmission protocol is experimentally verified.(3)On the basis of the above research,based on front-end and back-end technology and algorithm model deployment technology,a Beidou communication ship recognition system has been designed and implemented.The system includes a video monitoring module,a video image analysis module,a Beidou warning information sending module,a warning data management and interaction module,and integrates functions such as real-time ship monitoring,intelligent ship analysis,warning sending,and information interaction,The feasibility of the above target recognition algorithm and Beidou short message communication technology has been demonstrated.