Marine Image Target Tracking Research Based On The Probability Hypothesis Density

Nowadays,as an important part of computer vision,target tracking is widely used in traffic management,driverless,human-computer interaction and other fields.Target tracking is playing an important role in all aspects of artificial intelligence.China is a big ocean country.Artificial intelligence products are used in sea,such as unmanned boats,maritime traffic,maritime rescue and so on,with the gradual deepening of the development of the ocean.More and more occasions require target tracking technology to assist in processing.However,Most target tracking algorithms are studied for pedestrians or vehicles in terrestrial environments.Due to the complexity of video images,Maritime target tracking faces many problems.Therefore,this paper studies the ship target tracking problem in the marine environment.For this problems with target features and classic algorithms,this paper proposed an improved method based on GM-PHD algorithm to improve tracking performance for maritime targets,The specific work is as follows:(1)Due to target tracking algorithm is less used at sea,so this paper analyzes the characteristics of the marine environment and marine targets and we build an experimental platform to verify the tracking effect of the classic target tracking algorithm on maritime targets.With the impact of the marine environment on tracking targets,as well as the impact of the sea target itself,size,etc.this paper analyze these issues to propose corresponding improvement methods.(2)In order to solve the problem of the waves and wave caused by moving ships and unknown number of targets,this paper is based on the GM-PHD tracking algorithm to track sea targets.This paper is based on space-time correlation method to solve the problem of the waves and tail waves.Label the targets that appear in the scene to distinguish between different targets.This paper analyzes marine target motion model to solve the impact of waves on the movement of ships.Finally,The improved GM-PHD tracking algorithm is analyzed on the marine dataset to verify the feasibility of the improved method.(3)Aiming at the problem of poor tracking accuracy of weak targets at sea,the method of re-detection after target tracking failure is proposed.This paper analyzes the GM-PHD target tracking failure mechanism and the characteristics of maritime weak target matching and template matching.The SATD template matching method is used to re-detect the weak target tracking at sea,and the characteristics of the maritime target are improved.The improved SATD template matching algorithm is analyzed to verify the performance of the improved algorithm.(4)In this paper,the improved GM-PHD tracking algorithm and the improved SATD template matching method are combined,and the target frame approximation processing method for maritime target mutual occlusion is proposed.We give a specific algorithm flow,and the improved algorithm is tracked under the ocean dataset to verify the feasibility of the improved algorithm.The improved algorithm has a good tracking effect on the target in the marine environment,and improves the tracking success rate of the weak target at sea.The improve algorithm is compared with classic target tracking algorithm and finds that the tracking performance of the maritime target is greatly improved.