Laserfiche WebLink
MEMO 89.2 <br />possible errors associated with spectral similarities between irrigated crops and other land cover types <br />(e.g., dry-land crops, riparian vegetation, urban areas, etc.). <br />In the second step, reference data collected from FSA and other sources was used to `train' the ERDAS <br />Imagine MLCS and produce a per pixel classification of crop types. For use in validation and QA/QC, a <br />series of experiments was conducted to evaluate the MLC classification against other classification <br />techniques found in peer-reviewed literature (for references, see SPDSS Memorandum 89.1). Among the <br />techniques evaluated were `Unsupervised Isodata", `Hybrid Unsupervised/Supervised' as well as `simple <br />decision rule' classifications using NDVI imagery and Tasseled Cap imagery. When compared to the <br />other classification techniques, the MLC consistently provided better results. In general, these results <br />were 5 to 10% more accurate than the other classifications. The MLC classification was iteratively <br />refined through signature analysis techniques and input from water users and other sources. <br />In the third step, post classification rules were applied to assign crop types to previously defined irrigated <br />parcels. These rules included the derivation of zonal statistics, such as area tabulation of crops and <br />majority fraction, as well as a measure of homogeneity per parcel. Additional rules applied also included <br />the use of parcel types and shapes (e.g., center pivots). Figure 13 illustrates the rule-based classification <br />process. These rule-based classifications were performed by querying parcels within ArcMap using rules <br />described below. <br />'-~~ ~} ` MLC classification <br /> <br />'~!--' and derived GIS data <br />~`.: <br /> <br /> Parcel attribute <br />_ <br />~_ , ~~ Zonal statistics table <br />t table <br /> <br /> <br />i <br />' ~ <br />Maj. <br />Crop Crop "' Frac. Homog. Crop <br />Parcel Type <br />Parcel Yp Yp l <br />T e 1 T e 2 "' 1.0 7 <br />... <br />10012 Grass <br /> <br />10012 6 0 ... ... 0.5 4 <br />i <br />Classification <br />10013 Small Grains <br />10013 1 7 ... ,,, 0.95 g <br />i • <br />rules ~ 10014 Corn <br /> <br />... , <br />10014 3 2 10 $ <br />i 10015 Alfalfa <br />10015 4 0 ... " <br />... ~ ... <br />I <br /> Irrigated crops <br />Irrigated parcel classification <br />data (per parcel) <br />Figure 13. Rule-Based Parcel Classification of Crop Types <br />5 The Maximum Likelihood Classifier MLC is based on the probability that a pixel belongs to a particular class. The basic <br />equation assumes that these probabilities are equal for all classes, and that the input bands have normal distribution. The <br />maximum likelihood classifier uses the Gaussian threshold stored in each class signature to determine if a given pixel falls within <br />the class or not. The threshold is the radius (in standard deviation units) of a hyperellipse surrounding the mean of the class in <br />feature space. If the pixel falls inside the hyperellipse, it is assigned to the class. The class bias (BIAS) is used to resolve overlap <br />between classes, and weights one class in favor of another. <br />Page 19 of 45 ~Rfversfde FecAnotogy, fnc. <br />4'JaYer Resources Errgi~ecr:np an~i CansuFlrnp <br />