Your story:
Assume that as part of your GIS plan you need to provide information about the acquisition of remotely sensed data for the temporal updating of a variety of data layers in the local enterprise GIS. You've heard that new commercial remote sensing satellites now provide extremely high resolution (0.5 - 1.0 meters) imagery in the visible portion of the spectrum that can be used for applications that previously required aerial photography acquisition. This new generation of satellites has provided a new set of opportunities for organizations who want to update their geospatial information systems. This exercise will introduce you to the challenge of assessing the best option for the acquisition of remotely sensed data for integration into a GIS.
Questions to answer:
Prepare a write-up that outlines your assessment and provides an initial recommendation. Your write-up should include a map of the area and a full discussion of your findings including each of the factors listed above. A complete write-up will indicate a discussion of the logistical challenges, associated costs (on a per square mile basis), and a discussion of the options that are available.
Factors to Consider:
Raster or image data can be used as a source of vector data . There are several ways that raster data can be used to extract vector data or to adjust the vector data. The underlying principle seems to be that a picture is worth a 1000 lines. In other words, people trust pictures more than they do other forms of data. First, it must be remembered that distortion in images, even ortho-rectified ones, is most often greater than that achieved by a competent land surveyor. Secondly, the rendering of a line in an image is not to the same accuracy as it is in a vector due to pixelation, and for course rasters this is much worse of a problem. Thirdly, most raster do not have associated attribute data.
The ways to make vector data are many and will only be very briefly touched here. The most common today is heads up digitizing, i.e., tracing around the edges as they are seen on an image in an editing session. This is what Susan did for Montserrat. This can be quite accurate of it is done carefully and at the same scale as the image was taken, which is the raster resolution. Tests have shown that any form of manually digitizing has error and that rarely will two operators, or even the same operator if they repeat the same segment, end up with the same line work. However, with care and good work practices, this can be very nearly as accurate as the image (it is often hard to tell how accurate an image is so the final accuracy and precision is hard to judge definitively). In Spatial Analyst there is a menu choice to convert from raster to vector. Thus it does so in a "black box" manner. Care must be taken to look at the result critically. Often it is quite good, especially for simple raters, like a simple line drawing map. It is no use at all for actual photographs. Edge extraction can be used on these but is often so messy to be useless even as a drawing guide in heads up digitizing. There is an extension in ArcGIS called ArcScan that can be used to supervise how ArcGIS converts raster data. ArcScan allows you to trace lines manually and the program follows the lines in the raster for you. It has a number of rules about when to bridge a gap in the pixels and when not to jump across a gap; it lines up with the center of a line if it is more than few pixels wide. It also has many parameters to allow scanned material to be automatically converted after it has been suitably prepared. ArcScan was used to convert nautical charts to vector data. The last method is to use an image to spatial adjust vector data. This is the opposite of ortho-rectified Anchor points placed on lines and a linkage made to where the lines have to move to line up with the same visible feature in the image data. A note of caution here, it is necessary to ensure that the image is in the same projection, same datum and is error free -- otherwise you could be ruining a perfectly good dataset to agree to something just because it is pleasing to the eye. This problem often comes up when aerial photography is used as a backdrop to vector data. For example, bridges will appear to be wrong in an aerial because they are above the surface. It is worse if the overpass is high, and even worse when they are off to the edge of an aerial where parallax makes the distortion greater. It is rare that ortho-rectification is done so carefully as to remove all such artifacts or that a DEM is available that exactly follows the surface. Further, ortho-rectification is only as good as the DEM, the height control and exactness of the camera lens mathematical model. That is a long way of saying do not always trust your own or the camera's eyes.
More information on ArcScan can be found in the ESRI Helpfiles.
This module is one week in length. Please refer to the course Calendar tab, above, for the due date.
1. Readings:
Required:
Optional:
2. Post a project write-up including:
3. Discuss the weekly topic on the discussion forum.
4. Begin to finish writing your course paper. Your paper is due in the Dropbox of Lesson 5 Folder next week .
You have just completed module 4.
Don't forget...if you have any questions, feel free to post them to the Lesson 4 Discussion Forum.