The Moving Objects Of The Track Data Mining Research

In recent years, along with a variety of wireless communication technology (eg,Bluetooth, Wi-Fi, GPRS,3G, etc.) the rapid development, more and more mobiledevices are used in various applications. This makes a lot of mobile data containinglocation information (also known as trajectory data) is stored in a variety oflocation-based services applications, typically have a cell phone location-basedservices, GPS car navigation, wildlife tracking system. However, the current systemto provide location-based services provided by relatively simple functions, such asmobile location services only provide the location of the user, or according to theirgeographic location to provide further information around, and GPS car navigationsystems usually only in accordance with The target user input to provide a navigationroute from the nearest (this is often not the nearest path to the best people need thepath). Positioning data to accumulate gradually make people aware of: the effectiveconcentration of mining from the movement of mobile data for the furtherdevelopment of location-based services related to the application system has a pivotalrole. The complexity of the data path of traditional data mining techniques andmethods can’t be directly used in the field. Therefore, many domestic and foreignresearchers of mobile data mining algorithms in-depth study, four directions ofresearch mainly focused on mobile data clustering analysis, mobile data anomalydetection, frequent pattern mining and the location and trajectory prediction.The main content of this paper are clustering analyze of moving objects andtrajectory outlier detection of moving objects. Trajectory outlier detection of movingobjects is to find out trajectories which the moving mode of trajectory is different withtheir specific areas in a large number of trajectory data. Clustering analyze of movingobjects is mining clustering similar trajectories use of cluster analysis methods, so thatto find out the movement rules of moving objects and patterns of behavior. Attrajectory clustering analysis of moving object and outlier detection of moving objects,the trajectory of the similarity measure is the key technology. This paper analyzes thespace-time characteristics of the trajectory, the use of trajectory data and image data similarity are formed by the set of points, image matching in the pattern recognitionapplications to similar match of trajectory. But the trajectory data and image data arenot exactly the same, trajectory have some the characteristics that moving mode isdifferent from the image, so in order to can better application the distance of similarmatch to the trajectory, made the corresponding improvement for the distance ofmatch. Trajectory outlier detection of moving object use line Hausdoff distance assimilarity measure between trajectories. In order to be able to comply with Hausdoffline distance, this paper will divided trajectory into sub-trajectories which is linesegments composed of two points according to certain trajectory segmentationalgorithm. Finally, create a R-Tree index for each sub-trajectory to reduce the distancemeasure calculation between the sub-trajectories, thereby enhancing the efficiency ofthe algorithm. Cluster analysis of moving object have two work in the paper, the firstis to only consider the movement direction of trajectory, use the trajectory describedby the form of flow vector, and will Hausdoff distance measure turn into the form offlow vector. So with the direction of the track when trajectory clustering. The secondis to consider not only the direction of the trajectory also has the speed, in order toreflect the these two features of trajectory, this paper will has been improved forHausdoff distance. The process is fixed a trajectory and move one trajectory toeliminate public bias, and the basic comparison unit is applied when matchingtrajectory, the distance compute is the corresponding point to a point. To be able tomore quickly find the basic unit use the point distance matrix. Finally, we usetraditional clustering algorithms to cluster sub-trajectories.