1 / 11

MF Tree Comparisons

Manu Shukla 6/14/09. MF Tree Comparisons. MF Tree Algorithm. Boundary points of each map feature forms a curve that divides universe into 2 partitions The tree’s first level is a hash where the hash node for a map feature points to the

lise
Download Presentation

MF Tree Comparisons

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. Manu Shukla 6/14/09 MF Tree Comparisons

  2. MF Tree Algorithm • Boundary points of each map feature forms a curve that divides universe into 2 partitions • The tree’s first level is a hash where the hash node for a map feature points to the • set of cells in the spatial index on both sides of a partition and on the partition line • linear least square fitting curve • Overlaps, containment etc. are not an issue as the hash entries are independent • As we process a query • Refine the cells with the MF hash if a map feature present in query • test for distance etc. using the curve for the map feature • Query the spatial index to get a set of locations that match cells from MF tree and combine with other locations from query • Get documents from text and spatial index

  3. MF Tree Example • For the map features mf1, mf2,mf3, mf4, using their coordinates, build curves c1,c2, c3, c4 • Mf1 can be a lake, mf2 shoreline, mf3 a housing development and mf4 a highway • Build a hash where • Mf1 points to c1 and a set of cells c1s1 within mf1, c1s2 on the curve and c1s3 outside • Mf2 entry points to the cells in the two regions it partitions the universe to and set of cell on the boundary • Same for mf3 and mf4 • Using the sets of cells in both partitions and boundary and the curve we can filter cells based on query terms like within 5 miles of mf2 • Can access 1 or 2 sets of cells mf1 mf2 mf3 mf4

  4. Experiments • Built a hash (in java) with up to 1 million mf-tree elements based on a synthetic dataset • Each hash element had four fields, the cells in partitions, the curve and cells on the curve for each map feature • To index the elements time taken:

  5. Experiments • To query for elements in the hash: • The resident size of the process went to 307MB

  6. Issues • Computational and storage expense • Paging issues with storing pointers to cells in hash in querying • How to keep the resident process size within limits

  7. QH-Index • Similar index used by Dong Li, Yu-hui Peng and Jiang-long Yin from South China University of Technology in Quadtree and Hash Table Based Index Structure for Indexing the past, Present and future Positions of Moving Objects, CSA 2008 • Authors use a modified quad-tree by adding time and speed variables and using that in their QH index

  8. QH Index • Entry[n] is an array for storing moving objects, with each moving object's structure being (OID,X,V,t) • R defines region and consists of (xlt,ylt,xbr,ybr)

  9. QH-Index • HashTable has one way linked list to preserve the historical trajectories of moving objects • LeafNodeLink points to the lef nodes of the improved QuadTree which stores moving objects • LinkedList points to the head node of the Linked List

  10. Experiment Results

  11. Conclusions • We can extend MF hash to include the trajectories of objects within the map feature • Can be used to track storms, military objects, mobile users on highway etc. • Big issue is to scale to millions of mobile users

More Related