Detection And Tracking Of Qiantang River Tidal Bore Based On Background Difference And Particle Filter Model

Qiantang River tidal is known as "the best in the world tide," attracting a large number of domestic and foreign viewers every year.The scene of Qiantang River tidal makes people surpried.When People watch the Qiantang River tide,dangerous situation may be lead to the occurrence of some tragedy.The detection and track system of Qiantang River tide is established to avoid some disaster.This system can serve the society better,which has become an urgent problem to be solved.In this paper,the moving target detection and tracking algorithm is applied to detect and track the tidal bore of Qiantang River.The main contents include: based on the background modeling of Qiantang River tide detection is build by the optimal feature selection and support vector mechanism,and tracking the Tidal Tide of Qiantang River is based on color feature and particle filter model.The main work and innovation are as follows:(1)The method of optimal feature selection: The pixels of a video image can have different types of features,and there are differences between the features.Firstly,the statistical density(Gaussian kernel function)is used to compute the probability density of a certain feature of the pixel in the video frame,and the density estimation value of different interval is calculated.Combining with the new feature extraction method.The optimal feature set is selected to form the initialized feature template.(2)Construction of adaptive background model based on support vector machine: The Least Squares Support Vector Machines(LS-SVM)are used to estimate the important information of pixel features in nonlinear dynamic scenes,and train corresponding to the position of the template map pixel eigenvalues,construct the background model.A new real-time background model updating algorithm is designed to reduce the effect of surface fluctuation on the background model.(3)Particle filter tracking algorithm based on color feature: The color feature has the advantages of local occlusion,deformation insensitivity and easy operation.It can restrain the influence of the fluctuation of water surface,the size of tidal bore and the variation of river width on the tracking effect of Qiantang River tidal bore.The H features are separated in the HSV color space,and the information of the particles is used to describe the state of the tidal bore and the effect of the trace.The color feature information histogram distribution of the particle set is constructed,the particle swarm system is constructed,and the particle state is updated.And then the particle swarm is resampled to track the tidal bore of Qiantang River.The simulation results show that the number of pixels correctly detected by tidal bore detection is higher than that of the same algorithm,and the F-Measure value is 78.1%.The algorithm dealswith the average speed of the video,and the background model update parameters can deal with the effect of surface fluctuations.In the tidal bore tracking,the algorithm of the pixel error in about 7units,the average speed of video processing in 89.7ms / frame,can be a long time to track the steady flow tide,to meet the requirements of the subject.