Bentley Descartes - Raster Registration

Pascal Cloutier — Thu, 19 Mar 2020 12:39:41 GMT

Current Revision posted to ContextCapture | Descartes | Pointools | Orbit Wiki by Pascal Cloutier on 3/19/2020 12:39:41 PM

Raster Registration

With Bentley Descartes it's possible to register images so they can be correctly located to a given coordinate system. This is done by warping them to fit a reference or base document. The base document can be an image, a vector file, a drawing, a map sheet, or a list of known points.

Control Points

The first step in the register process is to place control points. Control points are locations on the image that typically represent a feature for which the world coordinates are known. This would include features such as the corners of a building, a road intersection, the edge of a property and so forth.

To obtain the world coordinates for a feature:

Locate the feature in the field and use a GPS to extract the latitude and longitude values.
Locate the feature in Google Earth and read the latitude and longitude values directly from this application.
Locate the feature on a georeferenced document such as another image, or MicroStation design file.

Note that the precision of the world coordinates are very important for the accuracy of the registration.

In preparation for placing control points, first define their locations by placing a marker at the location. For example, you can place red circles on the unregistered image feature to indicate that they are in an uncorrected location. Then place green circles on the corresponding locations using the read world coordinates obtained from the field, Google Earth or georeferenced document. The green markers represent the Base System while the red markers represent the Uncorrected System.

In the graphic below, the image needs to be registered to the vector map which is georeferenced correctly.

The Place Control Points tool can be used to position the points needed for Register to process the image. Pre-defining the locations of the points simplifies and streamlines this workflow.

To refine the registration and/or ensuring that the image lies accurately over the known features, you can move or delete any control point during the registration process.

Register

The register process requires you to select the model that could be best used to transform the unregistered image. This selection is made in the Register dialog which is also where you can manage the quality of the control points.

Control Point List: This is the list of control points that are currently defined in the model.
Base System Coordinates: The XY coordinates (2D) of the control points on the base system. This system is the one into which the image is to be transformed.
Uncorrected System Coordinates: The XY coordinates of the control points on the uncorrected image.
Model Selected: The model that should be used to transform the unregistered image. The models name is descriptive: (Minimum number of required points) Name (Transformations). The minimum number of points specify the minimum number of control points needed for defining the transformation. The Name describes the transformation. The Transformations describes the basic geometric transformations that the selected model can represent.
Residual: Used to specify whether the calculated and displayed residual values apply to the Base System or to the Uncorrected System. These values describe the performance of the selected model to transform the control points.

Xbcar = Xuc + Xr
Ybcar = Yuc + Yr

Where Xuc and Yuc are the uncorrected system XY coordinates of the control point (See #3). And Xr and Yr are computed residual of the control point for the X and Y coordinates respectively (See #5). Xbcar and Ybcar are the XY coordinates of the pixel at the control point's base XY coordinates (See #2) after being resampled .
Global Error Calculation Metric: The metric shown at the right of the list box that is used to quantify the global error of the residual of all control points. The metrics available are the Root Mean Square of all residuals or the Standard Deviation of all residuals.
Residual Type: The type of residual computed which can be Uncorrected System (See #5) and Base System, which is the residual related to the difference between the base XY coordinates (See #2) of the control point and those coordinates when passed in the inverse of the selected model.

Resampling

After defining the control points and choosing the correct model for the registration process, the last step is to resample the image which is done using the Resampling dialog.

Files Selection: Before the resampling can start, it's necessary to choose source, model and, if necessary, output files. The model file is creating from the register dialog by clicking on the menu File > Save As.
Window Selection: Specifies a window for the input raster and the output raster.
Resampling: Specifies common resampling parameters such as the type of resampling, pixel size and the background color, which is the color applied to destination pixels which have no corresponding source pixels.
Job Manager: Starts the resampling process. Note that many resampling tasks can be started as a resampling batch process.

Productivity Tip : Since the resampling can be time consuming and CPU instensive, it is advisable to group resampling tasks together and process them during off hours.