Difference between revisions of "XB PointStream"

From CDOT Wiki
Jump to: navigation, search
(Parser Interface)
(Parser Interface)
Line 47: Line 47:
 
<pre>
 
<pre>
 
/*
 
/*
   The parser needs to pass itself back to the library since the library could
+
   The constructor of the parser.
 +
 
 +
  @param {Object} obj - collection of named functions
 +
 
 +
  These functions pass the parser back to the library since the library could
 
   be working with many parsers simultaneously.
 
   be working with many parsers simultaneously.
  
  @param {Object} obj - collection of named functions
 
 
   start - must occur exactly once. Has one argument, the parser itself
 
   start - must occur exactly once. Has one argument, the parser itself
 
   end  - must occur exactly once. Has one argument, the parser itself
 
   end  - must occur exactly once. Has one argument, the parser itself

Revision as of 19:19, 28 December 2010

Releases

A cross-browser JavaScript tool which will emulate Arius3D's PointStream viewer. It will be able to quickly render a large amount of point cloud data to the <canvas> tag using WebGL.
XBPS 0.1
XBPS 0.2
XBPS 0.3
XBPS 0.4
XBPS 0.4.5 [Only Lib | Full]

XB PointStream Dev Links

3D Image Gallery
Blogs
Specifications
LightHouse
Github
Twitter
YouTube Videos

Development Resources

GPU Gems
GPU Gems 2

WebGL Specification
WebGL Cheat Sheet
OpenGLES 2.0 man pages

Occlusion Culling Algorithms


Parser Interface

There will be cases in which users have their own file format which they want to render with XB PointStream (for example, .ARA or .XML). We need to give users the ability to write their own parser and hook it into the library.

Users would write their JavaScript code which would implement the methods below. Once they do that, they would register their parser with the library by passing in an extension and their parser.

The library would then take care of the rest by creating an instance of their parser, call its methods and return a point cloud object.

/*
  The constructor of the parser.

  @param {Object} obj - collection of named functions

  These functions pass the parser back to the library since the library could
  be working with many parsers simultaneously.

  start - must occur exactly once. Has one argument, the parser itself
  end   - must occur exactly once. Has one argument, the parser itself
  parse - may occur one or many times. Has two arguments, the parser itself and a
  named collection of value types.

  See below for an example

*/
Constructor(obj)

/*
  Begins to load the resource.

  @param {String} path - path to resource
*/
load(path)


////// Getters

/*
  Get the version of this parser.
 
  @returns {String} parser version
*/
version

/*
  Get the number of points which have been parsed.

  @returns {Number} the number of points parsed so far by the parser.
*/
numParsedPoints

/*
  Get the total number of points in the point cloud, including points
  which have not yet been parsed.

  @returns {Number} the total number of points in the resource or -1 if unknown.
*/
numTotalPoints

/*
  Get the progress of the parser, how much it has parsed so far.

  @returns {Number} value between 0 to 1 or -1 if unknown.
*/
progress

/*
  The size of the resource in bytes.

  @returns {Number} the number of bytes in the resource or -1 if unknown.
*/
fileSize
/*
  The following example demonstrates how XB PointStream might use
  a particular parser.
*/

var parser;

function startCallback(parser){
  // started
}

function parseCallback(parser, attributes){
  parser.version;
  parser.numParsedPoints;
  parser.numTotalPoints;
  parser.progress;
  parser.fileSize;
}

function finishCallback(){
  // finished
}

// create a hypothetical parser and set the callbacks
parser = new XYZParser({ start: startCallback,
                         parse: parseCallback,
                         end: loadedCallback});
// load some resource
parser.load("pointcloud.xyz");

Performance Issue Ideas

  • Spatially partition point cloud into Bounding volume hierarchy and do bounding volume frustum culling
  • Do static rendering
  • Render low LOD when rotating point cloud
  • Push as much as we can to CPU and GPU
  • Use progressive "mesh" optimization to render objects if they are far from the camera
  • Allow user to control LOD
  • Stream point cloud, dynamically create and merge bounding volumes
  • Pre-create Bounding volume hierarchy