Trajectory Summarization Framework Based On Group Denoising

With the popularize of the surveillance equipment, global positioning system and mo-bile computing device, the trajectory data come from human, boat, vehicle and aerostat is increasing rapidly. The management of trajectory data, which includes trajectory summa-rization, clustering, and outlier detection, is extraordinary significant.This thesis proposes a novelty trajectory summarization framework. The massive com-plicated trajectories are summarized to a small quantity of abstracted trajectories. The pat-terns of the data are recognized and the innate characters of the data are highlighted by compressing the similar trajectories. There are three functions of the proposed framework,trajectory summarization, which produces the abstracted trajectory, trajectory outlier detec-tion, which recognizes the abnormal trajectory, and trajectory clustering.We propose the framework from the perspective of signal processing. That is, trajec-tories are designated as signals, manifesting the copious information that varies with time and space, and information denoising is exploited to concisely communicate the trajectory data. Basically, the method involves two components. Trajectories are,in a preprocess-ing step, resampled to have a same number of sample points, by investigating the Jensen-Shannon divergence (JSD) to measure the difference of the trajectories before and after being re-sampled. The resampled trajectories are matched into groups according to their similarity and,a non-local denoising approach based on wavelet transformation is devel-oped to produce summaries of trajectory groups. Our new framework can not only offer multi-granularity abstractions of trajectory data, but also identify possible outlier trajec-tories. We propose three quantitative metrics,namely,Granularity of Abstraction (GA),Degree of Redundancy (DR) and Fidelity (FID), to objectively and fully evaluate the pro-posed framework for trajectory abstraction. Extensive experimental studies show that the proposed framework achieves very potential results in trajectory summarization, in terms of both objective evaluation metrics and subjective visual effects.