Technische Universität Wien
Orientation and Processing of Airborne Laser Scanning data
Department of Geodesy and Geoinformation - Research Groups Photogrammetry and Remote Sensing
Detects gaps (i.e. void pixels) in raster models and provides several interpolation methods to close them.
The generation of Digital Terrain Models (DTM) usually leads to void pixels where buildings and vegetation were removed. Module FillGaps aims to close these gaps with qualified approximation of void pixel using information of valid neighbouring pixels. In a first step, connected void pixels get joined to gap objects with boundary information.
Each of the boundary pixels is adjacent to at least one void pixel and contains a valid pixel value. Based on the boundary information the void pixels will be filled with qualified values in a second step.
The following fill methods (parameter method) can be used to close gaps:
When using method feature the following statistic features (see parameter feature) of the boundary pixels are available to close gaps:
By default, the entire raster data set is filled. To fill only a part of the data set (e.g. along a flight strip), the fill area can be defined with the parameter fillMask as Shape-File (*.shp) or raster mask (e.g. *.tif). A good option is to use the module Module Bounds to determine the convex hull or the alpha shape of your dataset.
To detect outer boundaries of the dataset and avoid filling them with inappropriate values the parameter boundaryRatio is used. The boundaryRatio defines the minimum ratio of valid boundary pixels of a gap to the total number of boundary pixels, including the pixel at the edge of the picture.
\( \LARGE boundary\ ratio = \frac{valid\ boundary}{total\ boundary} \)
The boundary ratio describes how well a gap is described by bordering pixels. If a gap has, for example, 15 valid boundary pixel with 20 boundary pixel in total, the boundary ratio is 0.75, which means that the gap is 75% bounded. Only gaps with a boundary ratio greater than or equal to the given threshold boundaryRatio will be processed and filled.
Examples for boundary ratio (see figure below):
If method adaptive is chosen to fill the gaps, an adaptive plane is fitted for each void pixel. The plane fit is performed based on all valid boundary pixels of a gap, with an adaptive weighting (parameter adaptive.weightFunc) of the pixels based on the distance to the void pixel, where the plane is calculated. The default weighting function \(\frac{1}{d^4}\) is used for the inverse distance weighting (IDW4) of boundary pixels. This means that a boundary pixel which is relatively far away from the void pixel gets a weight close to zero, while a close-by boundary pixel gets a weight close to one. By using this method, two neighbouring void pixels have similar weights on the boundary pixels in the plane-fit, thus, the method always results in a smooth interpolation within the gap.
Note: Due to the adjustment of weights in the plane fit computation for each pixel, considerable performance losses are to be expected for large gaps. To overcome this issue a subdividing strategy has been implemented, where the gap is divided into a grid with cell size adaptive.gridSize (see Figure below). In a first stage of the subdivided fill method only the grid outlines are computed with adaptive weighting based on all gap boundary pixels. In the second step all cells of the subdivided gap are filled by using the grid pixels of a 3x3 cell neighbourhood, which have been computed in the previous step. The parameter adaptive.maxPixelRigorous can be used to specify the maximum pixel size of a gap to be filled with the rigorous adaptive fill using all valid boundary pixels.
Possible values: feature ......... Use a statistic feature of the boundary gap points for filling plane ........... Uses tilted plane fit to fill gaps adaptive ........ Uses adaptive plane fit with inverse distance weighting triangulation ... Uses triangulation of boundary points kriging ......... Fill gap with kriging interpolationThe different fill methods use the boundary points of a gap for the approximation of void pixels inside the gap.
Possible values: BoundaryMask ..... Write file with boundary pixels of individual gaps BoundaryValues ... Write file with boundary pixels showing their actual value MBR .............. Write minimum bounding rectangles of individual gaps VoidMask ......... Write file with void pixels of individual gapsWrite additional gap information to files. The outFile parameter is used as file name basis. Additionally, a postfix is added to the filename, based on the selected gap information (e.g. _boundary, _void, boundary_values, _mbr).
The data used in the following examples can be found in the $OPALS_ROOT/demo/ directory. As a prerequisite the data of (flyover.laz) are imported into an OPALS data manager and a raster DTM is calculated using the following commands:
The dataset contains a flyover over a highway, which needs to be filled with proper values.
1.8.17