Research On Target Detection Recognition And Fusion Tracking Method Of Wide Area Submarine

With the application of emerging technologies that do not rely on air propulsion device technology,integrated vibration reduction and noise reduction technology on submarines,the concealment of submarines is further enhanced,and the role of submarines in naval warfare is highlighted.Therefore,how to detect,identify and track submarines in time and effectively becomes an important issue in anti-submarine warfare.The use of sonar alone for acoustic detection can no longer satisfy the increasingly complex marine environment.At different stages of the movement of the submarine,using different sensor devices to detect it has become the focus of research.This topic considers how to identify and track a submarine in the course of the submarine target from the port departure stage to the deep diving stage to the patrol stage.Different sensors are used in different stages,covering different detection methods and detection principles.During the departure phase,the submarine target is docked at the port or sailing on the sea;the submersible phase uses the submarine to dive underwater;the patrol phase is when the submarine reaches the target sea area to patrol.For different stages,the main contents include:First of all,during the submarine's departure phase,the submarine is docked at the port or sailing on the water.Due to the distance limitation,it cannot be detected by means of sonar deployment.It can be imaged and reconstructed using space-based remote sensing satellites.The main problem to be solved at this stage is the search and detection of submarine targets on a complex sea surface.In this paper,through the acquisition of optical remote sensing images of the sea surface with targets,a combination of convolutional neural network(CNN)and sliding window detection is used.Submarine target detection and search.Because sliding window detection is a traversal search for the entire image to be detected,in order to find the target more quickly and efficiently,the framework of the Faster R-CNN algorithm is introduced to detect the target of the submarine target,and optimize the feature extraction network in it to To achieve better detection results.During the deep diving phase of a submarine,the submarine sails underwater.Because of the special nature of the surface medium,SAR,optical sensors,etc.cannot directly detect underwater targets through the seawater,while the submarine sailing underwater will form internal waves on the sea surface.Wake,which has a long duration and large size,can be imaged and detected by SAR,and indirect detection of submarines can be achieved.In order to better realize the internal wave detection,we can use hydrodynamic principles to realize the modeling and simulation of the submarine internal wave wake,model the linear rough sea surface,and use the dual-scale method to conduct electromagnetic scattering analysis on the large-scale sea surface,and then Perform SAR imaging simulation.In the detection,the ship's wake may interfere with the internal wave's wake,so the convolutional neural network is used to classify and identify the internal wave's wake and the ship's wake in order to better identify the submarine's internal wave.Finally,during the patrol phase,the submarine targets are tracked by the passive sonar arrays deployed in advance and the active sonar buoys dropped by air.In this paper,the unscented Kalman filter(UKF)is used to filter the submarine targets.In order to improve the tracking accuracy,While maintaining system robustness,the distributed track fusion algorithm is used to achieve the fusion of the active sonar and passive sonar,to obtain the fusion track,and to achieve the fusion tracking of the submarine target.