Research On The Method Of Maritime Target Tracking Based On Spectral Filter

Research in the field of maritime target tracking has important academic value and practical significance,which can be used to meet people’s ever-increasing intelligent needs for ocean monitoring,marine resource protection and transportation.The spectral filter tracking method relies on graph representation,and introduces graph signal processing and spectral graph theory,which preserves excellent translation invariance and rotation invariance of the graph.It focuses on the adaptability to the local appearance changes of targets,but there are still some problems.Therefore,considering the complexity and diversity of the marine environment,this article mainly carried out the following work based on the spectral filter.A method for maritime target tracking based on kernelized spectral filter is proposed.Considering the unstable performance of the spectral filter tracking method when the maritime target is affected by the wake and illumination,we model the spectral filtering as a tracking framework based on kernel regression.According to the theories of graph signal processing and spectral graph theory,based on the description of kernel regression on graph signals,we deduce and discuss the construction of the filter,the calculation of kernel matrix,the solution of the kernel regression model,and the prediction of the tracking position.With the help of the nonlinear model with strong learning ability,it can effectively improve the precision under complex tracking conditions.After the experiments on the maritime dataset,the proposed method shows better tracking performance,especially when the target is affected by waves or wakes.A method for maritime target tracking based on spectral filtering and blob analysis is proposed.Considering that the spectral filter tracking method is easy to track drift when the ocean background is complicated,we compensate and correct the tracking result by blob analysis.We first perform image segmentation optimized by the genetic algorithm on the target region tracked by the spectral filter method,which effectively improves the search efficiency of the optimal segmentation threshold.Then we use the scale-invariant feature transform method to detect the key blob in order to ensure stability and accuracy.Finally,under certain conditions,we find the blob closest to the target position and update the position to the center of this blob,which further optimizes the tracking performance.Comparative experiments were also carried out on the marine monitoring dataset.The tracking precision of the proposed method is improved,especially when the target is blurred in motion or the ocean background is complicated.