Research On Target Detection And Tracking Algorithm For Surface Movement

As the rapid development of science and technology,China's maritime transportation industry is moving in the direction of high-speed development,the number of ships and bridges are increasing every year.In recent years,the accidents of ship collision,bridge collision and ship hit the island are more and more frequent.The risk factor of continuous improvement also caused a large number of casualties and property losses,but also brought some ecological Environmental issues.Therefore,it is necessary to design a video surveillance system for the detection and tracking of moving objects on the water.The method of moving objects detection is background modeling in water based on complex scenes.Firstly,we analyzed the characteristics of the traditional mixed Gaussian distribution model and the discussed the influence of the update rate of the mixed Gaussian distribution model on modeling and noise through the experimental data.Then,an improved hybrid Gaussian model was proposed to overcome the limitations of the traditional Gaussian mixture model.We adopted the average of the multi-frame images to instead of the first frame initializing model.The update rate of the Gaussian model is given in the modeling phase.In order to reduce the noise,a larger value reduces the value of the update rate as the number of image frames increases.Finally,the moving target region extracted from the improved hybrid model is processed by morphological processing and the number of statistical connectivity domains is used to detect the complete moving target region.The detection of moving targets is the base of target tracking.Particle filtering is used to track the moving objects.Particle filtering can not only deal with linear systems,but also deal with nonlinear systems.For the complex environment of water,the color of the target is affected by the change of illumination,and the background also changes at any time.Therefore,the single target feature can no satisfy the moving target tracking in complicated scene.So this paper chooses the particle filter tracking method based on multi-feature fusion of color histogram and edge(hog)feature.Firstly,the theoretical knowledge of particle filter,prior knowledge of target,state transition model and state observation model are introduced.Then,the multi-feature fusion strategy of moving targets is used to track the moving target.Finally,the particle filter algorithm based on color feature,edge and multi-feature is analyzed experimentally.It is found that the multi-feature particle filter has good accuracy and robustness.In this paper,the detection and tracking algorithms of water moving targets are mainly studied.By analyzing the characteristics of commonly used detection and tracking algorithms,an improved hybrid Gaussian algorithm is selected for target detection and particle filter tracking algorithm based on multi-feature fusion.