Trajectory Data Compression And System Implementation

Thousands of taxis equipped with positioning devices generate a large-scale of trajec-tory data everyday, supporting data source for Location-based Services(LBS). We analyze and mine from taxi trajectory data to get information for people’s life, for example, under-stand the traffic rule, help itinerary planning, analysis people’s behavior amongst others. However, the rapid popularization of mobile positioning equipments and development of the taxi business lead to the rapid growth of trajectory data, which brings many challenges in data storage, real-time query and data analysis. The trajectory data compression tech-nology is a kind of effective solution to the rapid growth of trajectory data. We extract trajectory features data and mine the distribution of features. Then we propose trajectory information measure standards based on features distribution, and put forwards two trajec-tory compression methods, Information-Based Compression Approache and Rate-Based Compression Approache, which assure good compression rate and high compressed tra-jectory quality. In order to deal with the typical query, we propose a distributed storage for compressed trajectory, assure high query efficiency and high throughput.Main contributions of this paper are as follows:·Trajectory Features Mining Framework Proposed a framework for trajectory features mining. We extract velocity, direction and position features from trajec-tories, and analyze the distribution of trajectory features based on Markov Chain. Gaussian Simple Model(GSM), Gaussian Mixture Model(GMM) and Exponential Distribution are used to fitting the distributions of features. Experiments on real datasets prove the validity of the models.·Trajectory Compression Proposed that applying information to measure the de-gree of importance for trajectory points. Based on information and features, we proposed two Mixture Compression methods, Information-Based Compression Ap-proache(ICA) and Rate-Based Compression Approache(RCA). They archive good compression rate and ensure the quantity of compressed trajectory. Compared with two existed trajectory compression methods on the same true dataset, experiments results show that our algorithms have better performance.·Distributed Storage and Query for Trajectory Aimed at common queries on trajectory data, we propose a distributed storage system based on HBase, to sup-port the efficient query and high throughput. We design efficient storage model and corresponding query algorithms. Then based on this we put forward the optimiza-tion strategy of BloomFilter, which can prune needless trajectories to improve query efficiency. Finally, the experiment results prove our storage system is efficient.