Laserfiche WebLink
<br />GJ? ~8':} <br /> <br />class areas (fig. 8) were determined for that part of the image adjacent to <br />the Hayden Powerplant (fig. 1): <br /> <br />Class Color oode <br /> <br />Deciduous vegetation----------------- Light blue. <br /> <br />Wet/irrigated hay meadow------------- Red. <br /> <br />Rangeland---------------------------- Purple. <br /> <br />Dry1and agriculture------------------ Yellow/tan. <br /> <br />Water-------------------------------- Dark blue. <br /> <br />About 10 percent of the pixels remained unclassified, and accuracy of the <br />class areas was not evaluated, because of lack of ground-survey information <br />obtained during the same time as the imagery. <br /> <br />Despite the problems with classification of complex land-use signatures, <br />the Image-l00 System proved useful to resolve many of the problems confronted <br />in analyzing the August 24, 1975, image. The rapid operation of hardwired <br />functions proved ideal for initial selection and evaluation of spectral <br />signatures of different levels of land-use categories. This exercise <br />provided information on the land-use categories and the level of detail for <br />each category that could be defined using digital analysis. <br /> <br />A map showing major landmarks (roads and rivers) also was registered to <br />the Landsat image through a trial-and-error procedure to evaluate overlay <br />capabilities. A mask of the drainage-basin boundaries then was digitized. <br />Many landmarks were visible on the August 24, 1975, Landsat image, and it <br />proved easy to register a map to the image. This process requires <br />approximately 30 minutes using.the Image-100 System. <br /> <br />Land use on the August 24, 1975, Landsat scene also was mapped using the <br />IDIMS. The emphasis was on determining the types and detail of land uses <br />that could be identified with some confidence from the imagery. Two clas- <br />sification techniques were applied to the digital data using this system-- <br />unsupervised or clustering techniques (p. 46) and supervised classification <br />with a maximum-l ikelihood classifier (p. 46). <br /> <br />Primary emphasis was placed on the supervised-classification approach <br />using the IDIMS because of the spectral variabil ity present within a given <br />land-use class. Clustering was applied only to a small part of the imagery <br />to evaluate the correlation between the classes selected by this technique <br />and the actual land-use or land-cover classes as interpreted from 1 :24,000- <br />scale black-and-white aerial photographs. <br /> <br />A region displayed on figure 9, encompassing an area extending from the <br />Williams Fork Mountains on the south to the Elkhead Mountains on the north <br />and from Cedar Mountain on the west to the town, Hayden, Colo., on the east <br /> <br />17 <br />