Track Optimization And Classification Based On Semantic Analysis

In the 21st century,under the general trend of globalization of the world economy,the ocean situation has undergone tremendous changes in all aspects,and ocean shipping plays an important role in international trade.As an important part of big ocean data,track data mainly records the important information such as the moving trajectory,status,and laws of moving objects at sea.The track data itself has large volume,high redundancy,low value density and distribution.The unequal characteristics strengthen the necessity of trajectory optimization processing in the process of trajectory information mining and application.Track classification is a further application of track information mining on the basis of track optimization.Related research has important research value in the fields of scientific investigation and marine traffic safety,and has become a hot research object in recent years.The characteristics of the track data itself will reduce the accuracy of the track classification and increase the calculation cost.In order to get the standard,continuous and effective track data,the original track data needs to be cleaned and optimized.For trajectory optimization of semantic analysis,firstly,the mean value filter is used to detect the noise points in combination with the speed and time information in the process of ship navigation,and the spatial position of the semantic information is applied to replace the spatial position information of the noise data,so as to realize semantic correction and reduce the data loss caused by the elimination of noise data.The track data period is long and the amount of data is large.In order to save the calculation cost,the track needs to be segmented.In this paper,proposes a segmentation method based on staying area detection,mainly by setting the endpoint distance threshold and comparing the actual endpoint distance obtain the starting point and end point of the data-intensive area,and compare the time interval between the starting point and the end point with the time interval of the smallest dense area to obtain the set of staying areas.In order to verify the correctness of this segmentation method,In this paper,proposes a spatial semantics based on the shortest endpoint distance to verify the correctness of the track segmentation algorithm,mainly by measuring whether the endpoint distance is the smallest within the acceptable threshold when the endpoint passes a certain area.Finally,in order to verify the effect of this optimization algorithm,the trajectory optimization experiment is carried out through actual data.After denoising,segmentation,and compression processing,the trajectory optimization is finally achieved.Data classification is a commonly used and effective analysis tool for data mining,and track classification can identify unknown tracks,providing an effective data foundation for the actual application of tracks.The existing track classification mainly uses the space or time information of the track to achieve the classification,but does not consider other information of the track,such as speed,course,etc.,or ignore the correlation between features,resulting in low classification accuracy,low time efficiency and other issues.Aiming at the above problems,In this paper,the Support Vector Machine(SVM)based on principal component analysis to achieve track classification.First,on the basis of trajectory optimization,extract the important trajectory points of the trajectory,increase the trajectory information density,and extract the statistical information of the five elements of the trajectory data:speed,acceleration,curvature,rotation angle,and direction respectively to obtain 65 track features,in order to improve the classification accuracy and operation efficiency,Principal Component Analysis(PCA)is performed on the 65 features,and 39 main features are extracted,which reduces the impact of redundant classification features on the classification accuracy,improve the time efficiency of model operation.Finally,the SVM classification algorithm is used to achieve track classification,and a good classification result is obtained,which proves the effectiveness of this method.In order to verify the advantages of the algorithm after adding principal component analysis,it is compared with the SVM algorithm without principal component analysis,and finally verifies that the classification accuracy and time efficiency are higher than the SVM algorithm without principal component analysis.Finally,the effectiveness of the algorithm used in this paper is summarized by comparing with Density-Based Spatial Clustering of Applications with Noise(DBSCAN)algorithm in three aspects:geometrical characteristics of track,Dynamic Time Warping(DTW)morphology,and statistical characteristics.