/pointcloud-benchmark

Primary LanguagePythonApache License 2.0Apache-2.0

pointcloud-benchmark

This project contains the pointcloud Python package (in python/pointcloud) which is used to deal with point clouds in various Point Cloud Data Management Systems (PCDMS's). Concretely the different PCDMS's are:

  • Combination of LAStools tools (lassort, lasindex, lasmerge and lasclip)
  • MonetDB: Flat table model
  • PostgreSQL: Blocks model (PostgreSQL point cloud extension by P.Ramsey) and also a flat table model
  • Oracle: Blocks model and also a flat table model

Alternative storage models based on Morton space filling curves are also offered as alternatives in some cases.

The main goal of the pointcloud Python package is to be used for the execution of benchmarks for the several PCDMS's. However it can be used in general:

  • To load/prepare the point cloud data through the loader tool (for example: import point cloud data from LAS files to PostgreSQL)
  • To query/retrieve portions of the loaded data through the querier tool (for example: get the points overlapping a rectangular region)

Installation

In the install folder you can find details on the requirements of the pointcloud Python package as well as installation instructions for the several PCDMS's.

Architecture

The architecture of the pointcloud Python package is based on implementing, for each PCDMS, at least a Loader class (that must inherit from AbstractLoader and implement the missing methods) and a Querier class (that must inherit from AbstractQuerier and implement the missing methods)

The Loader classes are very generic so they can be used for the benchmark executions but also for other applications to load point clouds in the PCDMS. The Querier classes are tailored to execute the queries specified in the benchmarks (see queries folder) but some of them implement parallel methods that may be interesting in other cases.

Content

  • The install folder contains the installation instructions.
  • The python/pointcloud folder contains the pointcloud Python package, i.e. the Loader and Querier classes for the different PCDMS as well as the loader tool (run/load_pc.py) and the querier tool (run/query_pc.py).
  • The ini folder contains the initialization files that are required to use the loader and querier tools.
  • lasnlesc contains the C binary loaders for PostgreSQL and MonetDB
  • sfcnlesc contains C and C++ version of the Morton-based queries used in the alternative storage structures.
  • The queries folder contains XML files where the benchmark queries are defined
  • the doc folder contains a usage guide for the various PCDMS. It contains how to use them independently of the pointcloud Python package and also with it.

Execution

In order to use the loader tool of the pointcloud Python package execute the following command:

python/pointcloud/run/load_pc.py -i [init file]

In order to use the querier tool of the pointcloud Python package execute the following command:

python/pointcloud/run/query_pc.py -i [init file]

where [init file] is the initialization file for the related PCDMS that you are using (use the files in ini folder as templates)

Adding a new PCDMS

To add a new PCDMS into the pointcloud Python package it is required:

  • To add a Loader class. The new class must inherit from AbstractLoader and implement the methods initialize, process, close, size, getNumPoints.
  • To add a Querier class. The new class must inherit from AbstractQuerier and implement the methods initialize, query, close.
  • To add a ini file which must contain at least the following sections and options:
[General]
Loader: pointcloud.newpcdms.Loader
Querier: pointcloud.newpcdms.Querier
ExecutionPath: newpcdms
LogLevel: DEBUG
UsageMonitor: True
IOMonitor:

[Load]
Folder: 

[Query]
File: 
NumberUsers: 1
NumberIterations: 2