Research Of Taxi Abnormal Behavior Detection Based On Cloud Computing

With the development of sensor technology, communication technology, storage technology and computing power, more and more taxis equipped with GPS sensor devices, have generated a large amount of location data in their daily operations, which provide us with a good opportunity to analyze and mine valuable information. This paper uses the data to detect abnormal taxi behavior. The goal is to identify abnormal running track of vehicles automatically and determine whether the driver has intentional detour behavior. It can protect the interests of passengers and help regulate taxi service, which has practical significance. In this paper the main research work are as follows:(1) In order to achieve the target of detecting abnormal taxi trajectories, the paper presented relevant definitions about trajectory firstly, designed the overall detection system framework and analyzed data processing from offline and online processing stages.(2) To solve the problem of discontinuity after trajectories gridding, the paper proposed AE-AUG(Augmented method of angle and edge) trajectories completion algorithm, which could quickly find a path connecting two non-adjacent grids.(3) This paper proposed the s-iBOAT(iBOAT with State) algorithm to solve the critical problem of abnormal trajectory detection. The algorithm improves isolationbased online anomalous trajectory detection algorithm iBOAT(Isolation-Based Online Anomalous Trajectory Detection) through adding position of the newly detected point to each trajectory, which can simplify the process steps to find similar trajectories and enhance efficiency of the algorithm.(4) Processed the taxi GPS records to generate historical trajectories by using Hadoop platform. Combining the map gridding algorithm in Bing Maps Tile System, AE-AUG and s-iBOAT algorithm proposed in this paper, I implemented an abnormal trajectory detection system based on Web front-end technology.Tested the recognition effect of normal and anomalous child trajectories by siBOAT using detection system. The experimental result was good and consistent with the theoretical analysis. Detected all operations trajectories with same starting and end points. Analyzed two reasons how anomalous trajectories occur according to the overall detection. One is part of experienced taxi drivers who are familiar with the region select few and convenient paths which lead to anomalous identification. Second, others deliberately bypass when carrying passengers to take additional operating income. Studied how the abnormal judging threshold and grid size effect the detection of sensitivity, false positive rate and accuracy. The best values are 0.1 and 153 meters for them under the testing experimental conditions. Fixed the detection result based on the length comparison between anomalous trajectory and whole historical trajectories set. The experiment shown it can improve the detection accuracy and more suitable for the detection of real taxi detour behavior. Compared the implementation effect of detection algorithm before and after the improvement. The result shown s-iBOAT can increase the running speed and reduce response time while keeping the ability to recognize anomalous child trajectory and detecting whole trajectories accuracy...