Research On Fusion Detection And Situation Awareness Of Maritime Vessels

Territorial sea is about one-third of the land area in China.With the development of the country’s economy and society,maritime supervision is becoming more and more important.Perceiving ships at sea and supervising their navigation status is one of the important tasks of the maritime supervision department.Factors such as the response time,detection distance and detection accuracy of ship target detection will affect the accurate judgment of the ship’s navigation intention.It is difficult to meet the detection requirements by using only a single sensor.Using multi-sensor fusion technology can effectively solve this problem and improve the accuracy of Ship target detection capability in all aspects.Perceiving and predicting the sailing state of ships is another important part for ship supervision.The current supervision methods are mainly to perceive and monitor the current and past states of ship targets,and predicting the future state of ships will improve supervision capabilities.Obtain the data from various sensors,correlation fusion them that will obtain more information through reasoning and mining.To complete the perception of the current state of the ship and predict future intentions will provide technical support of supervision and early warning for the maritime management department.In this thesis that takes the ship target at sea as the research object,adopts multi-sensor detection technology to obtain the ship target track information,applies multi-sensor fusion technology to obtain multiple attribute characteristics of the ship target,and then combines environmental,meteorological,economic,political and other multi-dimensional data to introduce situational awareness.The technology obtains the current situation risk level of the target,and on this basis,the neural network is used to complete the prediction of the target situation level.The main research work of this thesisare as follows:1.The target tracking technology of high frequency ground wave radar(HFSWR)is studied,and a variational bayesian unscented kalman filter(VBUKF)tracking algorithm based on fuzzy joint probability data association(FJPDA)optimization is proposed.The systematic error of HFSWR is calculated,the VBUKF filter is designed to predict the track,and FJPDA algorithm is integrated to correlate the predicted value with the actual measured value,so that the track is updated in one step,and the next moment is entered,and the loop iteration is carried out until the termination condition is reached,so that the whole track can be tracked.The generated track is correlated with automatic identification system(AIS)track,which verifies the feasibility and effectiveness of the algorithm.2.This thesis studies the correlation between high frequency ground wave radar track,AIS track and spaceborne synthetic aperture radar(SAR)trace,and proposes a global optimal correlation algorithm with multi-threshold decision.The integration of multi-threshold decision can reduce unnecessary correlation track pairs before global optimal correlation,thereby reducing the calculation quantity of global optimal correlation and improving the calculation efficiency.The auction algorithm is used to solve the optimal solution and realize the correlation between HFSWR track and AIS track.The correlation between SAR trace and track realizes the correlation between the trace and the trace that makes up the track,and the multi-threshold global optimization algorithm is also used to realize the correlation of the three sensors.The correlation results show that compared with the nearest neighbor and the mean nearest neighbor,the global optimal algorithm with multi-threshold decision can obtain more correlated track pairs.Finally,according to the correlation results of the three sensors,a multi-feature data set is established with the ship target as the research object,which provides data support for the situation awareness of the ship target.3.With the support of acquiring multidimensional characteristic data related to the target,a fuzzy evaluation situation assessment algorithm based on subjective and objective combination weighting is proposed to realize quantitative situation assessment of the current state of the target.Taking whether the target is close to a specific area as a condition,five risk levels of the target’s sailing intention are defined,and the evaluation system and index are established by using the relationship between the target characteristics and the intention judgment,and the evaluation index is quantified.Then,the corresponding weights of evaluation indexes are obtained by the method of subjective and objective combination weighting,and the evaluation grade is determined by combining fuzzy functions.After calculating the membership degree of each evaluation index,the fuzzy vector is formed,and the multiplication operator is selected to complete the final evaluation,and the risk level of the target navigation intention is obtained.By comparing the real intentions in the data set,it is proved that the situation assessment algorithm is accurate and effective.4.Aiming at the "state" of the current target,a gate recurrent unit(GRU)neural network algorithm based on improved particle swarm optimization is proposed to predict the future "potential".In order to obtain better learning effect,the parameters of each weight should be debugged in the learning process of GRU network.In the process of calculating the optimal solution,particle swarm optimization of differential evolution algorithm is introduced to optimize the parameters of neural network,to predict the future intention level.In the process of prediction,358 samples are used as training set and 90 samples are used as testing set.By comparing the results,the correct rate can reach 96.7%,which can realize reliable prediction.Finally,a visual simulation platform is programmed and built in the PyCharm development environment,all algorithms above are embedded in it.The ship’s track and situation levelcan be dynamic show in visual display.