LSOBIA

OTB Module providing Large Scale Object Based Image Analysis functionalities.

Dependencies

OTB
MPI (developped and validated with mpich-3.2)

How to build it

LSOBIA can be built like any other otb remote module You can build it either from within OTB's sources or outside it.

How to use it

LSSegmentation

Parameters

LSOBIA provides an OTBApplication, LSSegmentation (Large Scale Segmentation).

Parameters: 
        -progress                          <boolean>        Report progress 
        -io.im                             <string>         Input image path  (mandatory)
        -io.out.dir                        <string>         Output directory  (mandatory)
        -io.out.labelimage                 <string>         Label Image Name  (optional, off by default)
        -io.temp                           <string>         Directory used for temporary data  (mandatory)
        -algorithm                         <string>         Segmentation algorithm name [baatz/meanshift] (mandatory, default value is baatz)
        -algorithm.baatz.numitfirstpartial <int32>          Number of iterations for first partial segmentation  (optional, on by default, default value is 1)
        -algorithm.baatz.numitpartial      <int32>          Number of iterations for partial segmentation  (optional, on by default, default value is 1)
        -algorithm.baatz.stopping          <float>          Value for stopping criterion  (optional, on by default, default value is 40)
        -algorithm.baatz.spectralweight    <float>          Value for spectral weight  (optional, on by default, default value is 0.05)
        -algorithm.baatz.geomweight        <float>          Value for geometric (shape) weight  (optional, on by default, default value is 0.95)
        -algorithm.baatz.aggregategraphs   <string>         Aggregation of graph traces [on/off] (optional, off by default, default value is on)
        -algorithm.meanshift.maxiter       <int32>          max number of iterations  (mandatory)
        -algorithm.meanshift.spatialr      <float>          Spatial bandwidth  (optional, off by default)
        -algorithm.meanshift.spectralr     <float>          Spectral bandwidth  (optional, off by default)
        -algorithm.meanshift.threshold     <float>          Threshold  (optional, off by default)
        -algorithm.meanshift.ranger        <float>          Spectral range ramp  (optional, off by default)
        -algorithm.meanshift.modesearch    <string>         Activation of search mode [on/off] (optional, off by default, default value is on)
        -processing.memory                 <int32>          Maximum memory to be used on the main node  (mandatory)
        -processing.maxtilesizex           <int32>          Maximum size of tiles along x axis  (mandatory)
        -processing.maxtilesizey           <int32>          Maximum size of tiles along x axis  (mandatory)
        -processing.writeimages            <string>         Activation of image traces [on/off] (mandatory, default value is on)
        -processing.writegraphs            <string>         Activation of graph traces [on/off] (mandatory, default value is on)
        -inxml                             <string>         Load otb application from xml file  (optional, off by default)

Monoprocessor execution

  otbcli_LSSegmentation "-io.im" "${INPUT_IMAGE}" "-io.out.dir" "${OUTPUT_DIRECTORY}" -io.out.labelimage "LabelImage" "-io.temp" "${TEMP}" "-algorithm" "baatz" "-algorithm.baatz.numitfirstpartial" "5" "-algorithm.baatz.numitpartial" "5" "-algorithm.baatz.stopping" "40" "-algorithm.baatz.spectralweight" "0.5" "-algorithm.baatz.geomweight" "0.5" "-algorithm.baatz.aggregategraphs" "on" "-processing.writeimages" "on" "-processing.writegraphs" "on" "-processing.memory" "2000" "-processing.maxtilesizex" "1000" "-processing.maxtilesizey" "1000"

Multiprocessor execution

Simply add mpirun -np ${NUM_PROC}

  mpirun -np 4 otbcli_LSSegmentation "-io.im" "${INPUT_IMAGE}" "-io.out" "${OUTPUT_DIRECTORY}" -io.out.labelimage "LabelImage" "-io.temp" "${TEMP}" "-algorithm" "baatz" "-algorithm.baatz.numitfirstpartial" "5" "-algorithm.baatz.numitpartial" "5" "-algorithm.baatz.stopping" "40" "-algorithm.baatz.spectralweight" "0.5" "-algorithm.baatz.geomweight" "0.5" "-algorithm.baatz.aggregategraphs" "on" "-processing.writeimages" "on" "-processing.writegraphs" "on" "-processing.memory" "2000" "-processing.maxtilesizex" "250" "-processing.maxtilesizey" "250"

This produces an image output containing labels of the Baatz Segmentation algorithm.

LSSmallRegionMerging

Other binaries

You can find useful application and binaries execution examples in the unitary tests. To enable these tests, simply build the module with BUILD_TESTING=ON.

Output Samples

Baatz segmentation

The next image is obtained with the previous Monoprocessor command ran on a sample image

TODO

Implement an application for small region merging algorithms
Implement an application for polygon simplification

Licence

Please see the license for legal issues on the use of the software.

pierre33 / lsobia Goto Github PK

lsobia's Introduction