CartographicDBValidation

Abstract

The aim of this work is, starting from the result of a detection (for example a road extraction), to enhance the results fiability by using a classifier fusion based validation.

Demspter-Shafer Theory

Explain: link between model fuzzy and descriptor + link to Dempster-Shafer

Application in cartographic database update: descriptors : NDVI/RADIOM/DBOVER/NDWI... link with the DS theory Format standard for the Hypothesis

Prequel: Road Extraction

-> Road extraction result on a Pleïade-like image

OTB Road extraction framework outputs inVD.shp

Fuzzy Model (requisite)

Perform the fuzzy model estimation (once by use case: descriptor set / Belief support / Plausibility support)

• Inputs:
  • a vector data of positive samples enriched according to the "Compute Descriptors" part
  • a vector data of positive samples enriched according to the "Compute Descriptors" part
  • a support for the Belief computation
  • a support for the Plausibility computation
  • a initialization model (xml file) or a descriptor name list (listing the descriptor to be included in the model)
• Output:
  • a FuzzyModel.xml file containing the model

Usage:

   ./otbDSFuzzyModelEstimation-cli -psin PosSamples.shp -nsin NegSamples.shp -BelSup "ROADSA" -PlaSup "NONDVI" "ROADSA" "NOBUIL" -DescList "NONDVI" "ROADSA" "NOBUIL" -out FuzzyModel.xml

FuzzyModel.xml contains the optimal model to perform the classifier fusion.

First Step: Compute Descriptors

The First step in the classifier fusion based validation is to compute, for every studied polyline, the choosen descriptors. In this context, the otbComputePolylineFeatureFromImage application can be used for a large panel of descriptors.

• Inputs:
  • an image (of the sudied scene) corresponding to the choosen descriptor (NDVI, building Mask...)
  • a vector data containing polyline of interest
  • a formula ("b1 > 0.4", "b1 == 0") where b1 is the standard name of input image first band
  • a field name corresponding to the descriptor codename (NONDVI, ROADSA...)
• Output:
  • a vector data containing polylines with a new field containing the descriptor value

Usage: To add the "NONDVI" descriptor to an input VectorData ("inVD.shp") corresponding to the percentage of pixel along a polyline that verifies the formula that have a NDVI > 0.4 :

   ./otbComputePolylineFeatureFromImage-cli -img NDVI.TIF -vdin inVD.shp -expr "b1 > 0.4" -field "NONDVI" -out VD_NONDVI.shp

where NDVI.TIF is the ndvi mono band image of the studied scene.

This step must be repeated for each choosen descriptor

   ./otbComputePolylineFeatureFromImage-cli -img roadSpectralAngle.TIF -vdin VD_NONDVI.shp -expr "b1 > 0.24" -field "ROADSA" -out VD_NONDVI_ROADSA.shp

   ./otbComputePolylineFeatureFromImage-cli -img Buildings.TIF -vdin VD_NONDVI_ROADSA.shp -expr "b1 == 0" -field "NOBUILDING" -out VD_NONDVI_ROADSA_NOBUIL.shp

Both NDVI.TIF and roadSpectralAngle.TIF can be produced using Monteverdi Feature Extraction capabilites, and Buildings.TIF can be generated using Monteverdi Rasterization module.

From now on, VD_NONDVI_ROADSA_NOBUIL.shp contains three descriptor fields. It will be used in the following parts.

Second Step: Feature Validation

The final application which, using the Dempster-Shafer theory, will validate or unvalidate the studied samples

• Inputs:
  • an enriched vector data VD_NONDVI_ROADSA_NOBUIL.shp
  • a support for the Belief computation
  • a support for the Plausibility computation
  • a fuzzy model FuzzyModel.xml
• Output:
  • a vector data containing only the validated samples

   ./otbVectorDataDSValidation-cli -in extractedRoads_enriched.shp -descMod FuzzyModel.xml -out validatedSamples.shp