Tangible Landscape plugin: Scanning
This is the description and explanation of tab Scanning of the Tangible Landscape plugin.
Rotation angle: the scan can be rotated by certain number of degrees around the vertical axis. This is useful when the orientation of the Kinect is different than the orientation of the physical model. For most setups 180 degrees is typically the right value.
The scanned point cloud usually includes all objects including model, sand pile, table, so we need to extract part of the point cloud which represents the physical model.
Trim vertically: Parameters to limit the point cloud vertically by two horizontal planes - top and bottom. The bottom value is the distance from the scanner to the table surface. See following figure.
Trim horizontally: Parameters to limit the point cloud by a bounding box - north, south, east, west. The values represent distances from the scanner to the border of the physical model. See following figure.
Calibration: There are 2 calibration phases. Before the first calibration phase, it is important to keep Kinect running for approximately 10 minutes to warm it up, because during this time the measurement is not stable. Press the main Start button and wait. Then stop scanning (using Stop button) and start the calibration. Calibration itself launches the scanning, so scanning must be off before that.
Calibration 1: The first phase of calibration procedure estimates the distance between the scanner and the table, and estimates how the scanner is tilted (the scanner axis should be completely perpendicular to the table, but often isn't). It is important to remove things from the table before calibration. Then press Calibration 1 and wait couple seconds. Then, Trim vertically, B field should be automatically adjusted and rotation matrix describing the tilting correction is saved into user settings.
Calibration 2: The second phase of calibration procedure is used when we need to find the boundaries of the physical model. Put the model on the table and align it with the projector's projection. It detects the model boundaries, so it is important to keep the surroundings of the model clear. Then press Calibration 2 and wait couple seconds. After that, the Trim horizontally, N, S, E, W fields are adjusted.
Trim tolerance: This is used for rectangular models to get rid of fuzzy edges. If the model is not rectangular, this field should be empty. The values go from 0 to 1, where high values trim the edges more.
The scanned model is scaled and georeferenced based on the provided information in these fields. If Reference DEM is provided, the scanned model will be translated and scaled to match the raster map given in Reference DEM. Similarly for Reference region.
Z-exaggeration is used when your physical model is vertically exaggerated to highlight terrain features. For example if the model is twice exaggerated, use value of 2, and that will vertically scale the scanned DEM to run further analysis on the not exaggerated terrain model.
There are several parameters which influence the quality of the reconstructed DEM.
Number of scans is the number of scans integrated into the resulting DEM. More scans can better capture details, but take more time.
Smooth value is the level of smoothness applied to the scanned point cloud to reduce noise. It can range between 5 to 15 (in mm), typically we use around 8-12, higher smoothing is better for water flow, lower smoothing for capturing finer details.
Resolution is applied when reconstructing the DEM into a raster. The value is in mm and should correspond to the actual spacing between points, which depends on the distance of the scanner. For most cases we use 2-3 mm. If it's lower than the point spacing, interpolation is needed.
Use interpolation instead of binning Binning is a fast way to reconstruct DEM as raster, each cell has the value of the point falling within that cell. This can result in empty cells. Interpolation is slower however it creates a continuous raster without empty cells and creates smoother surface as well.