TBC can generate elevation-based point clouds in the form of Digital Terrain Model (DTM) and Digital Surface Model (DSM).
DIFFERENCE BETWEEN THE DSM + DTM POINT CLOUD
Digital Terrain Model (DTM)
A DTM point cloud requires the least amount of time to create. It represents the bare terrain; it does not model elevated objects such as vegetation or buildings.
POINT CLOUD REPRESENTING DTM ELEVATION MODEL
A DTM point cloud has the advantage of representing the terrain with a reduced amount of points and does not include a point for each pixel of the captured image.
Digital Surface Model (DSM)
A DSM point cloud accurately describes the visible surface, including elevated objects such as
vegetation or buildings.
POINT CLOUD GENERATED USING DSM ELEVATION MODEL (INCLUDES ELEVATED OBJECTS)
The DSM point cloud creation process applies automatic image matching to every pixel of the captured image . This allows extreme height changes in the surface to be precisely monitored and measured. Complex terrain structures (rocks, embankment changes, terraces, trees, building sites, etc.) enable precise calculations.
DSM + DTM DIFFERENCE, IN DETAIL
The above to the left shows the denser DSM point cloud (blue circles) on top of the trees, which are needed to model the true position of the canopy.
The DTM elevation (red circles) is not sufficient to model the canopy correctly. Even if the DTM points are placed at the correct height position (yellow circles), it will not be the correct representation (blue dashed line) of the canopy.
Digital Surface Model - Highest Quality (DSM HQ)
TBC v5.60 and later offer the Digital Surface Model (Highest Quality) option, which uses Semi-Global Matching (SGM) algorithm to provide cleaner and sharper edges in both orthomosaic and point cloud. The method runs additional processes such as stereo-model selection, multi-point matching, raw point cloud filtering for facade corrections and edge enhancements, and disparity map improvements for uncertain areas to deliver the best possible results.
To illustrate these improvements, the graphic below shows how multiple images can measure the same corner of the building. Combining different stereo-models for the same reference image allows for a multi-point matching. Each stereo-model pair creates a probability of how accurate the point can be measured.
BUILDING CORNER MEASURED FROM MULTIPLE PHOTO STATIONS
THE PROBABILITY OF MATCHING THE REAL-WORLD POSITION FROM MULTIPLE PHOTO STATIONS IS SHOWN IN THE CYLINDER (PURPLE, YELLOW)
Combining two probabilities increases the accuracy and probability of the final matched point lying in the intersection of the two cylinders.
Therefore, it is very helpful to have a strong overlap of images in the flight mission when running the Digital Surface Model (Highest Quality). It is recommended that the flight mission include 80% in-flight overlap and 80% lateral overlap. This method creates a true orthomosaic, as detailed in the “Orthomosaics” section below.
The Raster DSM is an interpolated point cloud image created from the point cloud deliverable. One of the benefits of the Raster DSM is the interpolation of gaps. If the original point cloud wasn’t able to generate points for some areas due to insufficient overlap or very low texture or resolution, the Raster DSM can fill gaps. Some raster formats (like the inpho SCOP DTM in TBC) can remember if one raster point was interpolated from the previous point or if the interpolation was needed because there was a gap.
As the raster solution always interpolates, it is mainly used for DTM products. It is strongly recommended to set the raster density lower than the point cloud density.
NOTE: This article is from a larger workflow bulletin produced by Trimble.