Research On Tracking Algorithm Of Moving Targets In Sea Environment

Multi-target Tracking(MTT)is a technology that uses signal processing to estimate the status of targets in real time,under the conditions that,the number of targets in the surveillance area is unknown,the observation is noisy,and that the source is uncertain.Multi-target tracking is the prerequisite for accurate target identification and command decision making,and plays a very important role in radar signal processing system.Multi-target Tracking has been widely applied in military and civilian fields,in which the sea environment is a common application scenario but very difficult to study.In the military aspect,the research on the sea moving targets tracking can realize the sea monitoring,the threat assessment based on the situation of the target,early warning and defense,etc.In the civilian aspect,it can maintain navigation safety and facilitate the search and rescue at sea,etc.However,the complexity of the sea environment greatly increases the difficulty of target tracking,because radar echo contains a large number of sea clutter and its characteristics are varied,which are difficult to accurately analyze.As a result,the performance of target detection and tracking is reduced and the practical effect is not satisfactory.Therefore,this thesis designs a tracking framework by studying the multi-target tracking algorithm to solve the practical difficulties of tracking moving targets in the sea environment.The specific research works are organized as follow:Firstly,a tracking algorithm framework suitable for tracking moving target on the sea surface is studied.The framework mainly solves three practical problems:the prob-lem of sea clutter interference,the problem of real-time tracking due to the huge number of measurements,and the problem of track management.In order to solve the problem of real-time tracking caused by the large number of sea clutter and dense targets,we combine different types of tracking algorithms to achieve the superposition of their respective advantages,and an algorithm structure is established which combines the Gaussian mixture Probability Density Hypothesis(GMPHD)algorithm and the Multi-ple Hypothesis Tracking(MHT)algorithm.The feature aided method is used to help the tracking algorithm distinguish the target from the sea clutter,and the working prin-ciple of the feature is explained.We also propose some reasonable track management methods and feature feedback methods to solve the problems such as the target being temporarily submerged by sea clutter and the difficulty of target track initiation and ter-mination.As a result,we can solve the common difficulties of moving target tracking in the sea environment under the same tracking framework.Experimental results have demonstrated the effectiveness of our proposed algorithm.Secondly,based on the measured data of the radar staring mode,the application and effectiveness of the proposed tracking framework are studied.When the tracking framework is built,the application and effectiveness of the algorithm details need to be studied according to the echo data.The focus of the research is on the data pro-cessing and feature extraction.Firstly,we analyze the structure and characteristics of the staring data,and study the differences between the target and sea clutter.Then modelling the feature statistics of radar echo in time domain,frequency domain and time-frequency domain,and analyzing the target and clutter determination method in the three-dimensional feature space.Finally,we apply the three-dimensional features to the tracking framework and give reasons for the method.The effectiveness of our proposed tracking framework and feature modeling is verified by the measured data of CSIR.Thirdly,based on the measured data of the radar scanning mode,the application and effectiveness of the proposed tracking framework are studied.Different from the radar staring mode,the scanning data is sent into the tracking system in the form of frames,and a time interval exists between frames.We first analyze the structure and characteristics of the scanning data.The echo of the same target has the azimuth "scal-ability",which provides an additional information for feature modeling.And due to the existence of scanning time interval,the feature stabilities of target and sea clutter are different between frames.We establish four echo feature models based on the character-istics of the data,which are the echo amplitude,appearance,RCS and Intersection Over Union of detection box,and analyze the difference between the target and the sea clut-ter in the four-dimensional feature space.In the end,the principle of multi-dimensional feature-aided tracking is introduced in detail,including feature-aided clutter filtering,feature-aided algorithm efficiency optimization,feature-aided track management,etc.The effectiveness of our proposed algorithm framework and feature modeling is verified under the measured data from Naval Aeronautical University.