Laserfiche WebLink
<br />. <br /> <br />. <br /> <br />:) I', <br /> <br />Task 2 -Image Processing <br />Image Stratification <br /> <br />Processing Purpose: Reduce the study area to include only land areas that contain crop and <br />irrigated lands, address any clouds in image, account for physiographic differences in the <br />imagery that will affect classification. <br />Benefits: Reduces classification time, signature review and editing and improves classification <br />accuracy. <br /> <br />This project covers all or portions of eight Landsat scenes covering the entire western slope of <br />Colorado. The classification strategy should include steps to minimize the spectral variability's <br />introduced at this regional scale. Atmospheric, physiographic and phenologic factors all <br />contribute to this variability. Bauer et al. (1994) was able to increase overall classification <br />accuracy by 10 to 15% after subsetting imagery by physiographic region and classifying the <br />subset images separately. The contractor will perform atmospheric haze reduction and clip the <br />TM imagery to the division watershed boundary (preprocessing step). However, because the <br />State is interested in a fairly small group of agricultural land use classes, the contractor will <br />narrow the spectral variability further by extracting for processing only those land areas that <br />support the land uses which are being classified. Two agricultural masks will be prepared for <br />each image, the first will include those areas that are identified as agricultural field boundaries, <br />and the second will be defined as agricultural regions encompassing the fields but also <br />including any areas that appear to contain crops or irrigated acreage. <br /> <br />Clouds and cloud-shadows can be present in Landsat TM imagery. Clouds and cloud- <br />shadows will be digitized on each TM image and used to create mask files. <br /> <br />A final set of image masks will be developed based on physiographic region. Previous work <br />conducted by BOR divided the western slope into two regions (montane and arid) based on <br />interpretation of TM data and basinwide DEM data. The separation of the arid lowlands from <br />the higher elevation montane will reduce cross-strata signature overlap and within strata <br />variability which will improve classification. Physiographic stratification was a key element in <br />the LCRA project success. <br /> <br />Work Product: <br />Landsat image masks for physiographic region, agricultural region, fields, clouds <br />and cloud shadows. The masks will be used to increase classification accuracy <br />by splitting the Landsat data into more homogeneous areas, thereby reducing <br />spectral variability. <br /> <br />Image Enhancement <br /> <br />Processing Purpose: Perform Tasseled Cap transformation and Principal Components <br />Analysis for the Irrigated Acreage Status and Crop Type classifications, respectively. <br />Benefits: Reduce data redundancy, classification time, signature review and improve <br />classification accuracy. <br /> <br />Page 13 <br />