1 / 34

VLDB 2008

VLDB 2008. TRACLASS : Trajectory Classification Using Hierarchical Region-Based and Trajectory-Based Clustering Jae-Gil Lee, Jiawei Han, Xiaolei Li, Hector Gonzalez University of Illinois at Urbana-Champaign. Outline. Motivation TraClass : Trajectory Feature Generation

nigel-berry
Download Presentation

VLDB 2008

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. VLDB 2008 TRACLASS: Trajectory Classification Using Hierarchical Region-Based and Trajectory-Based ClusteringJae-Gil Lee, Jiawei Han, Xiaolei Li, Hector GonzalezUniversity of Illinois at Urbana-Champaign

  2. Outline • Motivation • TraClass: Trajectory Feature Generation • Trajectory Partitioning • Region-Based Clustering • Trajectory-Based Clustering • Classification Strategy • Performance Evaluation • Related Work • Conclusions

  3. Classification Classifier Unseen data (Jeff, Professor, 4, ?) Scope of this paper Features Feature Generation Prediction Training data Tenured = Yes Class label

  4. Trajectory Data • A trajectory is a sequence of the location and timestamp of a moving object Hurricanes Turtles Vehicles Vessels

  5. Trajectory Classification • Definition: The process of predicting the class labels of moving objects based on their trajectories and other features • Applications: Homeland security, weather forecast, law enforcement, etc. • Example: Detection of vessel types (e.g., container ships, tankers, and fishing boats) from satellite images

  6. Previous Studies • Several trajectory classification methods have been proposed mainly in the fields of pattern recognition, bioengineering, and video surveillance • A common characteristic of earlier methods is that they use the shapes of whole trajectories to do classification, e.g., by using the HMM Note: Although a few methods partition trajectories, the purpose of their partitioning is just to approximate or smooth trajectories

  7. Problem Statement and Observations • Problem Statement: Given a set of labeled trajectories, generate discriminative trajectory features that make a specific class distinguishable from other classes • Observations: (1) Discriminative features are likely to appear at parts of trajectories, not at whole trajectories; (2) Discriminative features appear not only as common movement patterns, but also as regions

  8. Motivating Example • Observation 1: Parts of trajectories near the container port and near the refinery enable us to distinguish between container ships and tankers even if they share common long paths • Observation 2: Those in the fishery enable us to recognize fishing boats even if they have no common path there Sub-trajectory Region

  9. Limitations of Earlier Methods • The classification accuracy of earlier methods might not be high since the overall shapes of whole trajectories are similar to each other • Our framework TraClass aims at discovering both region and sub-trajectory features Overall shape

  10. Overall Procedure of TraClass • Extract features in a top-down fashion, first by region-based clustering and then by trajectory-based clustering Trajectory partitions Recursively quantize non-homogeneous regions Region-BasedClustering Trajectory partitions in non-homogeneous regions Repeatedly find finer-granularity clusters Trajectory-BasedClustering Region-based and Trajectory-based clusters

  11. Our Contributions • Achieve high classification accuracy owing to the collaboration between the two types of clustering • Region features← Region-based clustering • Sub-trajectory features ← Trajectory partitioning and trajectory-based clustering

  12. Where We Are Now Trajectory partitions Recursively quantize non-homogeneous regions Region-BasedClustering Trajectory partitions in non-homogeneous regions Repeatedly find finer-granularity clusters Trajectory-BasedClustering Region-based and Trajectory-based clusters

  13. Class-Conscious Trajectory Partitioning 1. Trajectories are partitioned based on their shapes as in the partition-and-group framework [12] 2. Trajectory partitions are further partitioned by the class labels • The real interest here is to guarantee that trajectory partitions do not span the class boundaries Non-discriminativeDiscriminative Class A Class B Additional partitioning points

  14. Partitioning Condition • If the most prevalent class around one endpoint is different from that around the other endpoint, further partition it • Example: Prevalent class = Class B Need to be further partitioned Prevalent class = Class A Class A Class B

  15. Where We Are Now Trajectory partitions Recursively quantize non-homogeneous regions Region-BasedClustering Trajectory partitions in non-homogeneous regions Repeatedly find finer-granularity clusters Trajectory-BasedClustering Region-based and Trajectory-based clusters

  16. Region-Based Clustering • Discover regions that have trajectories mostly of one class regardless of their movement patterns • The region-based cluster is a set of trajectory partitions of the same class within a rectangular region regardless of their movement patterns (1) (2)

  17. Desirable Properties of Region-Based Clustering • Homogeneity: The class distribution in each region should be as homogeneous as possible • Conciseness: The number of regions should be as small as possible Note: Two properties are contradictory to each other • Need to find a good tradeoff between the properties conciseness homogeneity One large region Many small regions

  18. Translation into MDL Optimization • The minimum description length (MDL) cost consists of the description cost and the code cost • The former measures conciseness, and the latter homogeneity • The best hypothesis is the one that minimizes the sum of the description cost and the code cost • Finding a good quantization translates to finding the best hypothesis using the MDL principle

  19. Region-Based Clustering Algorithm • Progressively find a better partitioning alternately for the X axis and for the Y axis as long as the MDL cost decreases • Select the partition that has the maximum code cost and divide it into two parts in order to decrease the MDL cost (1) (2) (3) (4)

  20. Where We Are Now Trajectory partitions Recursively quantize non-homogeneous regions Region-BasedClustering Trajectory partitions in non-homogeneous regions Repeatedly find finer-granularity clusters Trajectory-BasedClustering Region-based and Trajectory-based clusters

  21. Trajectory-Based Clustering • Discover sub-trajectories that indicate common movement patterns of each class • The trajectory-based cluster is a set of trajectory partitions of the same class which share a common movement pattern (3) (4)

  22. Trajectory-Based Clustering Algorithm • Similar to our trajectory clustering algorithm [12], but incorporate the class labels into clustering • The algorithm is based on DBSCAN [5] • If an ε-neighborhood contains trajectory partitions mostly of the same class, it is used for clustering; otherwise, it is discarded immediately Non-homogeneousHomogeneous ε-neighborhoodε-neighborhood X O L1 L2

  23. Selection of Trajectory-Based Clusters • After trajectory-based clusters are found, discriminativeclusters are selected for effective classification • If the average distance to other clusters of different classes is high, the discriminative power of the cluster is high • Example: Class A Class B C2 C1 C1 is more discriminative than C2

  24. Generation of Cluster Links • A cluster link is a sequence of connectable (i.e., consecutive) trajectory-based clusters • Two clusters are connectable if they share enough trajectories (more formally, the ratio of common trajectories is higher than χ) • The benefit of cluster links is to derive also whole-trajectory features • Cluster links are added to the set of trajectory-based clusters for use in classification

  25. Classification Strategy 1. Partition trajectories by considering the class labels 2. Perform region-based clustering 3. Perform trajectory-based clustering 4. Select discriminative trajectory-based clusters 5. Find cluster links from trajectory-based clusters 6. Convert each trajectory into a feature vector • Each feature is either a region-based cluster or a trajectory-based cluster • The i-th entry of a feature vector is the frequency that the i-th feature occurs in the trajectory 7. Feed the feature vectors to the SVM

  26. Experimental Setting (1/2) • Use three real trajectory data sets • Animal movement data set • Movements of elk, deer, and cattle for the years 1993 through 1996 • Three classes: Elk, Deer, and Cattle • Number of trajectories (points): 38 (7117), 30 (4333), and 34 (3540) • Vessel navigation data set • Navigation paths of two vessels in August 2000 • Two classes: Point Lobos and Point Sur • Number of trajectories (points): 600 (65500) and 550 (125750) • Hurricane track data set • Atlantic Hurricanes for the years 1950 through 2006 • Two classes: Category 2 and Category 3 • Number of trajectories (points): 61 (2459) and 72 (3126) • Randomly select 20% of trajectories for the test set

  27. Experimental Setting (2/2) • Measure classification accuracy, training time, and prediction time for the three data sets • Compare two versions of the algorithm • TB-ONLY: Perform trajectory-based clustering only • RB-TB: Perform both types of clustering • TB-ONLY is expected to be no worse than earlier methods since it discovers also whole-trajectory features by cluster-link generation # of test trajectories correctly classified total # of test trajectories Classification accuracy =

  28. Overall Results • The classification accuracy of RB-TB is much higher than that of TB-ONLY • The training time of RB-TB is much shorter than that of TB-ONLY

  29. Features for the Animal Data Features: 10 region-based clusters 37 trajectory-based clusters Data: Three classes Accuracy = 83.3% Red: Elk Blue: Deer Black: Cattle

  30. Features for the Hurricane Data 1 region-based cluster 15 trajectory-based clusters These hurricanes entered the Gulf of Mexico and thus stayed longer at sea before landfall than others; They are likely to get strong because hurricanes gain energy from the evaporation of warm ocean water Stronger hurricanes tend to go further than weaker ones Gulf of Mexico Red: Category 2 Blue: Category 3

  31. Results for Synthetic Data • Effect of region-based clustering • Effect of the data size (scalability test)

  32. Related Work • Pattern recognition [1] e.g., speech, handwriting, signature, and gesture recognition • Classifying human motion trajectories • Employing the hidden Markov model (HMM) • Bioengineering [16] • Classifying biological motion trajectories • Video surveillance [15] • Detecting suspicious behaviors of pedestrians • Time-series classification [20,21] • Moving-object anomaly detection [14]

  33. Conclusions • A novel and comprehensive feature generation framework for trajectories has been proposed • The primary advantage is the high classification accuracy owing to the collaboration between the two types of clustering • Various real-world applications, e.g., vessel classification, can benefit from our framework

  34. Thank You!

More Related