[Return to the Software Reviews Index]
The Geographic Transformer (TGT)
reviewed by Sandy Prisloe
(reprinted from PE&RS August 1998)
Product Specifications Software:
Release Date: May 1998 Vendor:
RAM: 16MB
|
ASPRS does not endorse nor take any position concerning this
review. The views and opinions expressed herein are solely those
of the author and do not reflect the views and opinions of ASPRS.
Executive Summary TGT also can be used to register and transform scanned images of paper maps. This can be a particularly useful first step to prepare map data for on-screen digitizing. The reviewer has used this technique on a number of occasions and has found it to be an extremely useful and quick way to automate a variety of hardcopy maps. For geometrically correct image map data, TGT can register an image using an affine transformation with as few as three control points. For images that contain distortion, additional control points can be collected and images can be adjusted using a first or second order polynomial transformation. TGT uses what is essentially a two step process for preparing image data for use in desktop GISs. The first step involves registering the image by identifying pixel locations on the image and their corresponding real world coordinates. Once at least four registration points have been collected, TGT calculates the residual error associated with each point on the image relative to its pixel and real-world coordinates and the user has the option to use, modify, delete or ignore the registration point. During the registration and transformation process the user defines the input and output coordinate systems, a reference datum and units of measurement. TGT supports hundreds of predefined coordinate systems based on more than 30 map projections and allows users to create their own coordinate systems using a local datum if the need should arise. A number of industry “standard” image formats are supported including TIFF, JPEG, BMP, BSB Chart (input only) and uncompressed ADRG (input only). TIFF LZW compression, however, is not supported and it is recommended that simple flavors of image formats be used. Features The graphical user interface is organized to facilitate the
geo-referencing of image data in a step-by-step process. Across
the top of TGT window, see Figure 1, is a set of four drop-down
menus; File, Wizards, Options and Help. The File menu provides
basic functions which include opening source image files, opening
and saving reference files, opening and saving transformer setting
files, adding layers to a reference map window and managing layers
of data in a map viewing window. The Options drop-down menu provides two types of functions. One is to view and create linear, ellipsoid, datum and coordinate system definitions that can be used in the transformation process. The other function is to set preferences which include the cursor shape and color, coordinate format and precision, whether to show scroll bars in image display windows, setting mouse click properties, and other similar tasks. Preferences and coordinate system settings are saved from one TGT session to the next. The last drop-down menu is the Help menu that provides access to extensive on-line help that includes use of hypertext to define key words and concepts. Help information is organized through a Table of Contents that covers all the tasks commonly performed such as referencing, transforming, etc. The Help menu also has an “Obtaining Technical Support” choice that lists fax, e-mail telephone numbers and hours of operation that can be used to contact Blue Marble Geographics’ technical staff. Below the drop-down menus are four tabs titled Reference, Transform, Mosaic and View. Clicking on each of these tabs causes the lower portion of the application window to change to include information and functionality specific to the task being performed. The order of the tabs from left to right follows a logical geo-referencing sequence. When the Reference tab is selected, TGT window appears as in Figure 1. There are three smaller windows within the application window which are titled Source Overview Window, Reference Point Pick Window and Reference Map Window. The Source Overview Window is used to display the entire source image that is being geo-referenced and can be used to navigate around the image. When the image is first opened, a small blue rectangle appears centered on the source image and the area beneath the rectangle is enlarged and display in the Reference Point Pick Window. As the mouse is moved over either of these windows, its position in pixel coordinates is displayed beneath the image. Clicking the left mouse button when over the Source Image Overview Window repositions the image displayed in the Reference Point Pick Window to the mouse click location. Left clicking within the Reference Point Pick Window centers the image on the mouse click location and enters the pixel coordinates in the Reference Point List. Clicking with the right mouse button pops up a screen menu that zooms in or out on the click location by factors of 2, 4, 16, or 32 and also enters the location clicked on into the Reference Point List. By zooming in on the image in the Reference Point Pick Window, features on the image can be precisely located and their pixel coordinates can be stored in the Reference Point List. Once a feature’s pixel coordinates have been determined, it is then necessary to enter its corresponding real-world coordinates. TGT allows a great deal of flexibility in the format of real-world coordinates. Permitted are signed grid values, such as State Plane coordinates, and latitude and longitude which can be entered as signed decimal degrees; signed degrees and minutes; signed degrees, minutes, seconds; degrees and direction; degrees, minutes and direction; or, degrees, minutes, seconds and direction. Referencing an image involves collecting a sufficient number of pixel/real-world coordinate locations such that the image can be accurately transformed. The number of points required will vary based on the quality, scale, projection and geometry of the source image. The values for real-world coordinates can be typed directly into the Reference Point List or can be entered from an existing vector or image map that can be displayed in the Reference Map Window. This window supports the display of referenced images and vector data from ESRI shape files, MapInfo TAB/MIF files, DXF, DWG, and DGN files. The mouse functions the same way in this window to facilitate zooming in and out, panning and capturing coordinate values for posting to the Reference Point List. Left clicking on a location in the Reference Map Window automatically enters that location’s real-world coordinates into the Reference Point List. This function significantly increases productivity when already referenced vector or image data are available. As the Reference Point List is populated with reference point data, a set of residual errors is calculated and displayed to the left of the list. Errors are reported for source and reference Eastings, Northings, and Easting/Northing combinations. The user has the option of deselecting reference points, which marks them as not to be used in the error calculations and subsequent image transformation. As points are deselected, residual errors are recalculated so the effect immediately can be judged and a decision made whether removing the point will improve the registration/transformation. A reference point error graphing function also is part of TGT. By clicking on a button, the residual errors are displayed as circles on an XY graph. The circles are positioned at their respective image XY pixel locations and the size of the circles represents their error relative to one another. A set of radio buttons on the error graph permits selecting which residual error is to be displayed. It takes a little getting used to reading the error graph since the size of the circles is relative. For example, an image with four registration points and each with residual errors of 0.00 will display an error graph with four large red circles of the same size. The user also can select the type of transformation model to apply to the image with the set of registration points. Possible choices include affine (a minimum of three registration points are required), first order polynomial (at least four registration points are required), and second order polynomial (at least six registration points are required). The help files and user manual contain a brief discussion of which transformation is best suited to various types, scales and projections of source image files. Understanding the transformation models and their effects is extremely important. The same set of six registration points can produce quite different results when the same source image is transformed using each of the three transformation models. Once an acceptable set of reference points has been created, it can be saved as a Blue Marble Reference Setting File, an ESRI world file, and/or a MapInfo table. Images that require no transformation, such as USGS digital orthophoto quadrangles and that do not require being transformed into another coordinate system, can be used directly in other applications at this point. However, images that contain distortion or that need to be converted to another coordinate system need to be transformed. Clicking the Transform tab opens the Transformer Settings Window, see Figure 2. Within this window image input and output coordinate systems and other necessary parameters are specified. Using the Wizards drop-down menu will start a series of wizards to assist the user through this process. However, once familiar with the different steps, users can more quickly transform a number of registered images into a common output coordinate system without wizards. By default, the image that has just been registered will be used as the source image, along with its corresponding reference settings file, although other images and previously saved reference settings files can be selected through a standard Windows browse function. Clicking on the Reference Coordinate System button opens a dialog window in which choices for coordinate system, datum, units of measurement and other appropriate parameters are selected to describe the values of the real-world registration points. In a similar way it is also necessary to select destination coordinate system parameters by clicking on the Destination Coordinate System button. The user also must enter the pathname of the output image that will be created, must select an output reference file type (ESRI world file, MapInfo table, Blue Marble reference settings file or all three) and must calculate and enter an output resolution for the image. The output resolution is calculated as a function of the scale of the input image, the dots per inch at which it was scanned, and the desired output resolution in ground units per pixel. After establishing the image input and output coordinate system parameters, the user must define the area of the output image. Essentially, this permits cropping of the input image. Again, a coordinate system can be selected to help define the output image’s geographic area and coordinates for the northwest and southeast corners must be entered. Alternatively, buttons exist to select the entire area of the source image or to calculate a rectangle based on the geographic extent of the set of reference points. A last option that can be set is to tile the output image. Two tiling schemes exist; one tiles images based on tiles of 1.5' minutes of latitude and longitude whereas the other creates tiles based on a user defined tile size and row column origin. Tiling is particularly useful where a large image will be registered and transformed into a collection of smaller images with referencing for use in a project database. Parameters that are set for the transformation process are stored in a transformer settings file and can be restored at a later time for modification and use in performing subsequent source image transformations. Once all information has been completed in the Transform window and saved, the transform process can be initiated by pressing the Transform button. This creates a new output image, or set of images if tiling has been specified, while leaving the original source image unchanged. While tiling can be used to create a set of small easily displayed images, the Mosaic function, accessed by the Mosaic tab in TGT’s main window (see Figure 1) serves the opposite role of assembling a larger image from a number of smaller ones. With Mosaic, the user can create a geo-referenced output image from up to four smaller geo-referenced images that either abut or overlap each other. It supports the same image file formats as Register and Transform. However, input images need to have the same number of colors, the same color palettes, must be in the same coordinate system and have the same pixel resolution. Lastly, TGT also provides a convenient window within which to display geo-referenced image and vector data. This window is opened by clicking the View tab on the main window. Data layers can be added by clicking the File drop-down menu and selecting Add layer to view map window. Layers can be any of TGT’s supported image and vector file formats. When image files are added, the user also is prompted for the name of a reference file to use with it. There also is a View map properties, accessed through the File drop-down menu, that allows layers to be turned on and off, that permits changing the order in which layers are displayed and let’s the user center the display on a particular layer. While the mouse is over the View window, it functions in the same manner as when referencing images. Left mouse clicks re-center the display at the location clicked, unless this function has been disabled through user preference settings under the Options drop-down menu. Right mouse clicks permit zooming in and out by factors of 2, 4, 16 or 32. Performance Documentation The software also includes a Lotus ScreenCam movie that provides a tutorial on how to register and transform an image. The tutorial shows mouse movements and selections as though someone were actually registering an image. Where appropriate, text boxes pop up with explanations to further explain what is being shown. Technical Support Software Installation Ease of Use Pros The ability to get reference coordinates from data displayed in the Reference Map Window, which supports a wide variety of commonly used data formats, significantly improves the productivity of the software. Users of earlier version of TGT will be particularly pleased to have this functionality in the basic package. As reference points are added, the residual errors for each point are calculated and optionally displayed in a graph. This facilitates selecting which reference points to use to get the best geo-referencing and transformation. In setting up the parameters for a transformation, the user specifies the northwest and southeast corners of the output image. This allows the creation of a geo-referenced output image that can be significantly smaller than the source image and that covers only the geographic area of interest. The reviewer found this capability particularly useful when preparing images for small project areas. Cons Determining the output resolution for a transformed image requires that the user know the original scale and DPI of the image. From these data an output resolution can be calculated. It would be easier and more efficient if these values could be entered into an on-screen form and TGT calculated and suggested an output resolution which could then be modified by the user if necessary. TGT does not have a print function except for printing a graph of residual errors. It would be a valuable addition to be able to print a table of control points, reference settings, transformer settings, and mosaic settings and to be able to print images with control points as located during image geo-referencing. This would be valuable as a means to document what settings were used for a particular geo-referencing transformation process. It should be noted that this information is captured by TGT in a number of settings files but these easily can be “lost” or deleted once an image has been processed. Recommendations ………… Reviewed by Acknowledgments |