Print this page

filtering algorithm

The content is based on the Laserdata LIS documentation.

LIS DESKTOP offers different options for filtering the point cloud:

  • Filter LIDAR Points with Grid Surface and Filter LIDAR Points with Grid Surface (PC). Those modules need as input data an approximate bare-earth surface model. Based on the distance from this surface the classification of point cloud is performed.
  • Morphological Filter – the module provides morphological filter operations (erosion, dilation, opening and closing). Erosion is commonly applied to reduce, dilation to enlarge the size of features.
  • Progressive Morphological Filter (module LIS Filtering DT) – the module uses and adaption of the algorithm proposed by ZHANG (A Progressive Morphological Filter for Removing Nonground Measurements From Airborne LIDAR Data)
  • Remove Outliers (PC) – the module removes outliers from a point cloud.

From set of modules presented above two were chosen to filter the test point cloud. A Progressive Morphological Filter was used to create the bare-earth grid raster (the option of setting maximal height difference between iterations was deciding – the Morphological Filter did not have this option). Afterwards the Filter LIDAR Points with Grid Surface (PC) module has been used.

technical details

Progressive Morphological Filter works on a raster, which is created from the point cloud. The raster cell values correspond to the minimum height value of the point cloud within the raster cell.

The morphological operation of opening (erosion followed by dilation) is used to filter the grid surface. This operation is repeated iteratively – each time with a different window size and a different height threshold:

LIS Desktop equation 01
  • b- the radius of the initial windows size,
  • k – number of the iteration

LIS Desktop equation 02

  • dh0 – the initial elevation difference threshold,
  • s  - the terrain slope [rise/run],
  • c - grid cell size,
  • dhmax - the maximum elevation difference threshold.

The progressive morphological filter belongs to the group of slope-based algorithms. The filter works in the filter window, which moves along the raster. Firstly, the lowest value from the initial grid, within the window, is chosen and assigned to the new grid (erosion). Secondly on the ‘eroded’ raster the next step is performed – within the window the highest value is chosen and assigned to the new raster (dilation). Afterwards the values from so created raster are being subtracted from the initial grid. All difference values, bigger than the elevation threshold, are being rejected and removed from the initial grid (marked orange). So modified initial grid might be called the first intermediate DTM and is used as input for the second iteration, where a bigger filter window is applied. As a result of the second iteration a new intermediate DTM (second) is created. After the last iteration the final DTM is generated and used for the point cloud classification.

The example of operations using 3x3 filter window are presented below:

    LIS Desktop filtering 01    LIS Desktop subtraction

After subtraction the values greater than the elevation threshold are chosen (green). Based on those values the original heights (orange) are marked as objects and are not considered in the next iteration.

    LIS Desktop threshold    LIS Desktop threshold result

example of the filtering workflow

Beside the main LIS Filtering DT module additional five should be used:

  • Import\Export LAS -> Import LAS Files,
  • SAGA – Grid- Spline Interpolation -> B-Spline Interpolation (from Grid) (filling the holes in the raster grid),
  • LIS Point Cloud Tools DT -> Aggregate Point Cloud to Grid (creating a raster grid),
  • Import\Export LAS -> LAS Info,
  • Import\Export LAS -> Export LAS Files.

The first experiences with the filtering workflow might be hard, therefore a small summary of all steps has been prepared and is presented below:

LIS Desktop agregate

LIS Desktop agregate result

LIS Desktop filter

LIS Desktop fill the holes

LIS Desktop fill the holes 2

LIS Desktop classification

LIS Desktop export result

It is  important to precisely prepare the export of data (for each *.las file the parameters can be found in the Import\Export LAS-> LASInfo module).