e f tutorial atur ap feature pin mapping · 2003. 10. 28. · space used for identification of...

FeatureMapping

with

TNTmips®

TutorialFEATURE

MAPPING

Feature Mapping

Before Getting StartedThis booklet introduces techniques for interpreting surface features from images.Feature Mapping helps you interactively identify sample areas of the features youwish to map. It next searches for and highlights all the areas in the image that maycontain these features. You then use the interactive tools to decide whichhighlighted areas should be retained. This booklet will lead you step by stepthrough a series of exercises to familiarize you with all the basic tools in thepowerful Feature Mapping process in TNTmips®.

Prerequisite Skills This booklet assumes you have completed the exercises inthe Displaying Geospatial Data and Navigating tutorial booklets. Thoseexercises introduce essential skills and techniques, such as how to select and viewraster, vector, CAD, TIN, and database objects stored in Project Files. You shouldknow how to pan and zoom display objects and how to use the standard File /Object Selection process. This booklet does not present these basic skills again.

Sample Data The exercises presented in this booklet use sample data distributedwith the TNT products. If you do not have access to a TNT products CD, downloadthe data from MicroImages’ web site. The first set of exercises uses objects foundin the SECT27R Project File in the FEATMAP data collection in the LITEDATA directory.The second set of exercises uses objects found in the BEREA data collection (alsoin LITEDATA) in the BEREAPCA and BEREAMSS Project Files.

More Documentation This booklet is intended only as an introduction to thefunctions in Feature Mapping. Consult the TNT online reference manual, whichcontains more than 50 pages on Feature Mapping, for more information.

TNTmips and TNTlite™ TNTmips comes in two versions: the professionalversion and the free TNTlite version. This booklet refers to both versions as“TNTmips.” If you did not purchase the professional version (which requires asoftware licens key), TNTmips operates in TNTlite mode, which limits the size ofyour project materials and does not allow export.

Feature Mapping is not available in TNTedit, TNTview or TNTatlas. All exercisesin this booklet can be completed in TNTlite using the sample geodata provided.

It may be difficult to identify the important points in some illustrations withouta color copy of this booklet. You can print or read this booklet in color fromMicroImages’ web site. The web site is also your source of the newest GettingStarted booklets on other topics. While there, you can download an installationguide, sample data, and the latest version of TNTlite.

http://www.microimages.com

Merri P. Skrdla, Ph.D., 21 May 2002© MicroImages, Inc. 2002

Feature Mapping

Welcome to Feature MappingPhotointerpretation is a primary reason for convert-ing remotely sensed imagery to digital form. Clas-sification of the information contained in the imageryis one of the major applications of image interpreta-tion. The human brain is a tremendously powerfulanalytical tool that automatically classifies much ofthe information received without conscious effort.This interpretation is, however, limited to threecomponents translated into a screen image in thevisible portion of the spectrum. People can alsolearn to interpret visible representations of otherregions of the spectrum, such as color-infrared pho-tography that shifts and records photoinfrared, red,and green radiation into the RGB color space.

Remotely sensed images often contain many morebands than can be translated into a single RGBrepresentation, all of which may be important indistinguishing one ground cover type from another.Automated image processing lets you make use ofall of this information. TNTmips offers automaticclassification (both supervised and unsupervised),Feature Mapping, and manual interpretation. Auto-matic classification methods are designed for usewith high quality imagery that has uniform proper-ties throughout, such as satellite imagery. Aerialphotography and particularly aerial videographycontain artifacts not found in satellite imagery, suchas vignetting. Video images also contain significantnoise. Feature Mapping was developed specificallyto assist in classification of these lower qualityimages. The goal of Feature Mapping is to identify,mark, and measure features in a set of processingrasters by combining your knowledge of the studysite with TNTmips’ processing power.

STEPS

Vocabulary: Darkening atthe edges and corners of anoptical image is called vi-gnetting. The radiometricproperties of a feature at theedge or corner of an imagemay, thus, be different thanat the center.

Pages 4–10 provide a simpleexercise to acquaint you withhow this process works.Pages 11–14 introduce addi-tional Feature Mapping con-cepts using the samesample data. Pages 15–26introduce a more complexexample and additional fea-tures of the process.

launch TNTselect Process /Raster /Interpret /Feature Mapfrom the main menu

Feature Mapping

PreviousLayer Controls*

Skip Ahead

Analysis and Reference RastersVocabulary: Analysis ras-ters define an n-dimensionalspace used for identificationof cells that “match” samplecell values. Reference ras-ters are used to create thedisplayed image and may ormay not be included asanalysis rasters.

You are prompted to select analysis and referencerasters when you launch the Feature Mapping pro-cess. The number of raster objects that can beincluded for analysis is not limited by the FeatureMapping process; practical considerations deter-mine the optimal number of rasters to use. Thegreater the number of rasters to be analyzed, thelonger the time required for the classification stepthat identifies the prototype cells that fall within therange defined by the sample cells.

The analy-sis rastersare not usedfor display

STEPS

unless alsoselected forreference.

select the RED, GREEN, andBLUE objects from theSECT27RProjectFile, thenclick [OK]checkthat theSingleradiobutton inthe Se-lect win-dow ison, andclick [OK]selectSCREEN (from thesame file as theanalysis rastersclick [OK] tocomplete objectselection

View-in-View

Previous ZoomZoom Out

Zoom to Active Layer

Zoom InZoom 1X

Select Point

Select Area

Region-of-Interest

Categories

Define Samples

Draw Prototypes

Mark Features

Draw Marked Features

Protect Areas

Attach Vector

Calibrate

Classify

Feature Mapping toolbar

Full

RedrawStop

Add Layer

Zoom Box

Pan View

Select

GeoToolbox

contin-uedbelow

Clear Samples &Prototypes

Mark All

*Access tothe LayerControls letsyou use anynumber ofadditionalobjects, in-cluding vec-tors, for refer-ence.

Feature Mapping

Feature Mapping View WindowVocabulary: A samplepoint is a cell you identify asrepresentative of a particularfeature class. Cells identi-fied by the computer that fallwithin the range of valuesdefined by the sample pointsare called prototype fea-tures, or simply prototypes.Once you confirm that a pro-totype belongs to the sameclass as the sample points, itbecomes a marked feature.

The reference raster is automatically displayed atFull View, and Region of Interest drawing is theselected action once the analysis and reference ras-ters are chosen. We will save discussion of Regionof Interest drawing for later and move directly todefining samples. A low quality video image is usedto map water features in this first Feature Mappingtask. Remember that the red, green, and blue com-ponents of the video are selected for analysis, butyou are viewing an 8-bit composite color rasterobject for reference. The values of the cells in thereference image are not even considered when thecomputer searches for prototype features.

The FeatureMapping win-dow has bothmenus and atoolbar. Mostof the toolbaricon functionsare duplicatedon the menus.This bookletusually refersto the iconswhen both areavailable.

The toolbarprovides some display-related features not found onthe menus, such as the ability to add a reference layerand open the layer controls. The primarily red iconsnear the middle are the tools, most of which arefamiliar from all display processes. The Select Pointand Select Area tools work in tandem with theselected action. The actions make up the group oficons to the right of the tools. In some cases,changing the selected action changes the selectedtool. The last group of icons are commonly usedFeature Mapping functions (see facing page).

STEPSclick on the DefineSamples iconturn on the Options /Show Values check but-ton

Feature Mapping

Defining Sample CellsSTEPS

click on Zoom 1X icon

When the Show Values option is toggled on, thevalue for the cell beneath the cursor in each process-

sample range areshown in blackand those out-side the sample

ing raster is shownto the right of thecurrent range ofvalues. Cell valuesincluded within the

position the cursor overthe teardrop shape waterfeature that crosses theroad near the tophold down theleft mouse button anddrag within the confines ofthe water feature until anumber of cells are de-fined assamplesrelease themouse, click onClassify,and lookfor cells withinthe feature notincluded in thesample range(not redrawn inprototype color)click the leftmouse button ifany are found(cursor valueshown in red)

range are shown inred. Cells includedin the sample set areredrawn in yellow.

If you accidentallyinclude sample

points outside the water, you can remove them fromthe sample set by holding down either the or key and clicking on the cell again.

The value range needed to identify all the cells inthis feature as prototypes is Red 4–11, Green 9–14,and Blue 15–22.

Feature Mapping

Identifying Prototype Featuresclick on the Classify icon

click on the Full Viewicon to see the fullextent of prototypefeatures

You probably noted when you choose to definesamples that an additional layer appeared in theLegendView, which shows the sample and proto-type colors cur-rently selected. Asyou add featureclasses, their as-signed drawingstyles will also beadded to thislayer’s legend.

After you click onthe Classify icon,the message “Pro-cessing… Pleasewait…” appears inthe message line.The prototypecells are drawn in red after the processing is com-plete. These cells are the ones with values in the threebands selected for analysis that fall within the rangedefined by the sample cells (which, in this example,means a red value from 4 to 11, a green valuefrom 9 to 14, AND a blue value from 15 to 22).The computer has identified the cells thatsatisfy the sample range you established; itis now up to you to decide how many of theseprototypes actually belong to the featureclass you are mapping.

You can see that the ponds and standingwater were very effectively identified fromjust a few sample points of that feature type.When feature class identification is not so thorough,you can add more sample points after the classifica-tion step and classify again.

Many extraneous cells have also been identified asprototypes. We will discuss methods for handlingthis situation later. For now, let’s mark all featuresso you can save your progress.

Note: You can also initiateclassification by pressingthe right mouse button overthe viewing area when inDefine Samples mode.

Feature Mapping

Adding a Feature ClassSTEPS

click on the MarkFeatures iconclick on [Add] in theFeature Class Selectionwindow that opens*enter the name for thefeature class (water) inthe Add Class windowthat opensclick on [OK] to completenaming the class

When you are done identifying prototypes for aparticular feature class, it is time to mark the features.A feature map has two components: the raster objectthat shows all cells belonging to the same featureclass in the same color and an associated style objectthat specifies the color and the name for the class.This style object is the same as that used for assign-ment of display styles for vector, CAD, and TINobjects. These styles will remain associated with thefeatures if you vectorize your feature map. Marking

need to create a feature class for marked featureassignment before any features are marked. Youcan add more than one class by repeating thesteps on this page. A different color selection is

offered for each class as it is added. You can acceptthis color or adjust it if you have a particular colorscheme in mind. You can also open a feature classstyle object created in a previous feature mappingsession to provide the same color scheme and classnames if you are marking the same types of featuresin a new map.

The Feature Class Se-lection window is onlyopen when Mark Fea-tures or Draw MarkedFeatures is the selectedaction.

adjust the color used formarked feature display tocyan (100% green, 100%blue)click on [OK] to completecolor selection

features is a sequential, exclusive process. Oncea feature is marked as belonging to a class, itcannot be assigned to another class unless it isfirst unmarked (removed from the initially as-signed class).

When you click on the Mark Features icon, theFeature Class Selection window opens. You

* If you have used theFeature Mapping processbefore, the last usedstyles will be shown.Choose File/New, then re-turn to step 2.

Feature Mapping

Marking All FeaturesThe Mark All function is a shortcut for converting allprototypes to marked features. The alternativemethod is to click on each feature you want to mark

click on the FullView icon and notethat all prototypes, notjust those in the previ-ously visible area, havebeen marked

Mark All converts all prototype features, not just thosevisible in the view window, to marked features.

click on the Zoom1X iconclick on the Mark Allicon

with the SelectPoint tool or todraw a polygonwith the SelectArea tool and markthose prototypesthat fall at leastpartially withinthe area inscribed.Because featurescan be unmarkedlater, marking allfeatures and sav-ing the featuresraster is a quickmeans to save yourprogress to this point so you can exit the process andresume later.

Prototype features are redrawn in the color assignedto the selected feature class as they are marked so youcan watch the progress toward completion.

You can use the Mark All function wheneither Mark Features or Draw Marked Fea-tures is the selected action, and you have afeature class selected. This option is avail-able on the Special menu as well as from anicon.

You probably noticed that the water featureclass was added to the Features layer in Leg-endView. If LegendView is not on, chooseLegendView / Show / Left (or Right).

Feature Mapping

Saving ResultsIn the exercises so far in this booklet, you havecreated two different objects for the water featuresthat can be saved. The raster containing markedfeatures and the style object containing feature classinformation can be saved for further use in theFeature Mapping process or in other processes suchas Spatial Data Display or raster-to-vector conver-sion via automatic boundary tracing. Next you willlearn how to use the mouse to mark and unmarkfeatures, how to create regions of interest, generatea report, mark features in multiple classes, and createcategories. Before moving on, let’s consider savingthe objects created thus far.

All activities that involve saving or opening rasterobjects, in this case the features raster, are handledfrom the File menu in the Feature Mapping window.Feature style objects are saved from the File menu inthe Feature Class Selection window. A features styleobject is also automatically saved as a subobject ofthe features rasters to which it corresponds. Thepurpose of saving style objects separately is so theycan be used for subsequent feature maps of imagerywith the same types of ground cover. It is unlikelythat a more advanced user would save a style objectfor a single feature class, but to get familiar with theprocedure, go ahead and save this one.

Only marked features can be saved, so using theMark All function is a good strategy when you’reready to quit for the day without first deciding whichof the prototypes identified by the computer is amember of the desired class. You can exit the FeatureMapping process now if you want to take a breakfrom this tutorial.

to save the featuresraster created, selectFile / Features / Save,save it in the SECT27RProject File or in a newProject File; click onthe New Object iconin the Select Objectwindow and accept thedefault name

STEPS

to save the features styleobject, the Feature ClassSelection window must beopen, which means theMark Features or DrawMarked Features actionmust be selected (if youforget to save this or anyobject in the Feature Map-ping process, you will beasked if you want to saveit when you exit)choose File / Save As inthe Feature Class Selec-tion window and name anew object in the samefile as above

Feature Mapping

Resuming the ProcessSTEPS (if you exited)

launch the Feature Map-ping processselect the same analysisrasters selected previ-ously (see page 4)select the 8-bit compositecolor raster (SCREEN) fromthe same file for the refer-ence image

choose File / Features/ Open and select thefeatures raster savedon page 10

To resume the process, choose Feature Mappingfrom the main menu, select the same rasters foranalysis (the select process should default to thisfile), and choose either the same or another raster(s)for the reference image. To get your saved featuresraster back into the process for further manipulation,choose Open from the Features cascade on the Filemenu, and select the saved features raster. Thefeatures raster initially comes in as a hidden layerbecause the Region-of-Interest tool is ac-tive, but is automatically shown when youchoose the Mark Features tool.

The one thing lost when you Mark All, Save,and then resume is the identity of the samplepoints, because they were marked features at the timeyou saved. If you find you need to define moresample points in order to identify all members of afeature class after saving and resuming, you canreturn to defining samples for the class. If you usedMark All as a shortcut for saving your progress ratherthan because all prototypes were features you ulti-mately want marked, it is best to finalize marking ofthe prototype features before defining new samplepoints.

click on theMark Featuresicon

Feature Mapping

Unmarking FeaturesFeatures can be unmarked using either the SelectPoint or Select Area tool. You can unmark allfeatures in the selected class by drawing a polygonaround the full extents of the rasters being classifiedand clicking on the Exclude button in the Mark /Unmark Features window. You might want to unmarkall features if you had marked them with hole filling

turned off (discussed later) or if you usedMark All to save prototypes as we did.

Individual features are unmarked as a whole;you cannot unmark part of a feature. Featuresthat are even partly within the area inscribedby the Select Area tool are unmarked whenyou click on the Exclude button, as is anyfeature you shift-click on with the SelectPoint tool.

STEPSclick on the Select Pointtool and position themouse cursor overa feature you want tounmarkhold down the keythen click the left mousebutton

click on the SelectArea tool and drawa polygon that includes atleast part of a featureyou want to unmarkclick the Exclude buttonclick on the FullView icon if youhave zoomedindraw a poly-gon outsidethe extentsof the refer-ence rasterclick on[Exclude]

Polygon drawn tounmark featuresas shown below.

Feature Mapping

Marking Features with the Mouse

one at a time.

The Select Areatool marks all pro-totype featuresthat fall within thepolygon you draw.A partial featurecannot be marked;all contiguousprototype cells aremarked when anypart of the featureis included. Cellsthat fall within theSelect Area poly-

click on the SelectArea tooldraw a polygon that in-cludes some prototypefeaturesright-click over the view

or click on the In-clude button in theMark / Unmark Fea-tures window

click on the Select Pointicon and position thecursor over a proto-type feature and click theleft mouse button

STEPSWhen creating the sample set, only the Select Pointtool is active; both the Select Point and Select Areatools are available when marking features. TheSelect Point tool is initially chosen when you electto Mark Features. This tool lets you mark features

gon can be included in the selected feature class byclicking on the Include button in the Mark / UnmarkFeatures window or by right-clicking when the mouseis over the image.

Feature Mapping

Generating a Report

FT Unclassified 1 4524323.00 0.00 4524323.00 0.00 95.3 234256 0FT water 5 222183.47 63612.91 44436.69 12722.58 4.7 11504 14238

TotalArea

Total BoundaryLength

AverageAreaClass

TotalNumber of Cells

Total Numberof Boundary

CellsPercentof Area

Average Boundary Length

Number ofFeatures in Class

FS water 1 334 93 13171.86 3667.89 682 821FS water 2 233 188 5909.96 1652.70 306 370FS water 3 113 216 52822.65 16987.62 2735 3810FS water 4 162 250 19004.57 5460.14 984 1223FS water 5 221 378 131274.43 35844.56 6797 8014

Cells inFeature

BoundaryLength

BoundaryCellsArea

FeatureCentroid

line columnFeatureNumberClass

Type ofRecord

FD SUMP 703 Section 27 18 August 1991 17 May 2002 4746506.47 245760 100.0

A description of the contents of each fieldin every record type in the report is in-cluded when this button is on.

mark five features asshownselect Report from theFile menuenter summary data infor-mation and click [OK]click on the NewFile icon in theSelect File windowname the report fileuse Edit / Text Files (oranother text editor) toview the report

STEPS The Feature Mapping process provides a reportabout all marked features and the unclassified areaas well. Using Mark All with the features identified

by the sample points describedearlier results in 1972 features.Five features were marked for il-lustration in this exercise.

Total areaprocessed

Number ofcells in areaprocessed

Percentage oftotal rasterprocessed

1

3

2

4

5

Feature Mapping

Defining a Region of Interest

click on the Regionof Interest iconchange the line drawingmode to Stretch inthe Region of Inter-est windowposition the mouse cur-sor at the road intersec-tion at the upper left ofthe reference rasterclick and hold the leftmouse button as youdrag the cursor to theright, following the roadrelease the mouse buttonto insert vertices as nec-essary to follow the roadturn right at each inter-section and continuedrawing until you are nearenough to the start thatthe closing segment fol-lows the road wellright-click or click on theInclude button

STEPSYou can define a region of interest at any time in theFeature Mapping process by changing to the Regionof Interest tool. A region of interest is saved (File /Region of Interest / Save) as a binary raster, which canbe used as a mask in any TNTmips process.

All features and prototypes are hidden from viewwhen you select the Region of Interest tool so youhave an unobstructed view of your reference rasterwhile drawing.

Areas included in the region of interest are drawn justas before while areas excluded are drawn at halfnormal intensity. The defined region of interestlimits the area considered for any subsequent classi-fication activities and for reports generated.

areaexcluded

areaincluded

Features afterMark All without(right) and with

(below, right) re-gion of interest.

Note that markedfeatures are absent

along the edgesthat were excluded

from the region ofinterest.

Feature Mapping

Filling HolesWhether to leave the holes in prototype fea-tures when marking depends largely on thepurpose of the feature map. If you are mappingthe number of acres planted with a particularcrop, you will likely want to fill the holes that

may arise from emergent weed patches. If you wanta report that details the land area devoted to themature crop, the holes should remain when thefeatures are marked.

A hole can be any size or shape. Whether a hole isfilled depends on the extent to which the hole issurrounded by prototype cells. Hole filling requiresprototypes on eight sides (includes diagonals). Soif there is a diagonal gap at even one point betweenthe hole and the prototype, the hole will not be filled.Hole filling is on by default.

prototypes

hole filling off hole filling on

additionalboundarycell required forhole filling

this hole willbe filled

this hole willnot be filled

change to Full view ifzoomed in, and clickon Mark Featuresdraw a polygon aroundthe full extents of the ref-erence imageclick on the Exclude but-ton in the Mark / UnmarkFeatures windowmark a feature with andwithout hole filling

STEPS

Feature Mapping

More Complex Feature MappingThe number of analysis rasters is not limited inFeature Mapping, but the time for the classificationstep increases with the number of rasters. You shouldconsider dimensional reduction techniques for multi-temporal or hyperspectral imagery.

The remaining exercises make use of MSS imagerycollected on six different dates throughout the grow-ing season to create a crop map. Two of the fourcomponents from each date (Bands 5 and 7) wereused to generate a Transformed Vegetation Index(TVI, Process / Raster / Combine / Predefined). Thesix resulting rasters were then subjected to PrincipalComponents analysis (Process / Raster / Combine /Principal Components). The first four principalcomponents account for 93 percent of the totalvariance in the six-dimensional space defined by theTVI raster objects. The dimensionality has thus beenreduced from 24 to four for analysis.

select File / New to startanother feature maplocate and select the fourprincipal componentsgenerated for Berea(BEREAPCA Project File) foranalysisclick on [OK] to completeselection of rasters foranalysisclick on the Red-Green-Blue radio button in theSelect windowselect band 6 (NIR_6) asred, band 5 (RED) asgreen, and band 4 (GREEN)as blue (JULY30 folder ofthe BEREAMSS Project File)click on [OK] to completereference raster selection

STEPS

Applying some method that reduces the number of raster objects in a set is known asdimensional reduction. The intent is to retain most of the information in the original set by

The reference image is color-infrared produced fromthree of the original MSS bands. None of thesebands are used directly as analysis rasters.

some analysis or combination pro-cess. You might, for example, retainonly one raster from the bands thatshow a high degree of correlation.You can also use some other combi-nation method, such as generating avegetation index or using PrincipalComponents Analysis to retain the

information of interest infewer raster objects.

Feature Mapping

with the cursor over theView, type a then a, and scroll to the topmiddle of the rasterclick on theDefine Sam-ples iconturn on the ShowValues optiondefine samplecells on the right(darker) half ofthe center-pivotirrigated field (redcircle) at the topof the referenceimage until thesample rangeincludes all val-

STEPS There are a multitude of crop types on both dry andirrigated land in this 7.5' MSS scene of the BereaCreek West map quadrangle in Box Butte County ofwestern Nebraska. Using a color-infrared image forreference introduces a problem we didn’t encounter

Define Samples and Classify

ues (cursor valuesshown in black for othercells in this part of thefield)select Colors from theOptions menu, click onthe Prototype button, anduse the sliders to changethe color to magenta(100% red, 100% blue)click on [OK] to completecolor adjustmentclick on the Classifyicon

when working withthe video image. Ifwe leave red as the

fields. The colors forsample cells, prototypes,and protected areas, as wellas the display backgroundcolor can all be set from asingle menu selection.

color assigned to prototype fea-tures, it will be very difficult todistinguish prototypes from un-classified cells in most irrigated

Feature Mapping

Dealing with Messy PrototypesVocabulary: A pivot is thecircular crop area irrigated bya single center pivot irrigationsystem.

Ground truth information for the crops makes it clearthat many areas would be erroneously labeled ascorn if we used Mark All. Clearly, features should bemarked individually, which poses no problems inareas where extraneous prototypes aren’t contigu-ous with legitimate corn fields. The pivot where thesamples were defined has only a few extraneouscells, and they are not touching the prototypes wewant marked as corn in the right half of the pivot. Sothere’s no problem marking this feature; just click onthe prototype with the Select Point tool and MarkFeatures selected. Successful marking of other pro-totypes requires the introduction of two new skills:protecting areas and drawing prototypes.

STEPS

click on the MarkFeatures iconselect New from the Filemenu in the FeatureClass Selection windowadd a feature class andname it “corn”position the Select Pointtool over the prototypefeature at the top centerof the reference raster(right half of pivot) andclick

corn

soybeans

sugar beets

wheat

crop map for pivots in area outlined

Feature Mapping

Clearing Prototypes

select Clear Prototypesfrom the Special menu

You’ve marked one feature as corn, and now it’s timeto take the steps necessary to mark the other featuresin the same area cleanly. Additional sample cells areneeded to identify all pivots or partial pivots grow-ing corn. But first let’s take the steps necessary tokeep the corn separated from other crops in the areaoutlined on the previous page. The first techniquewe will use is protecting areas to prevent markingcells of other classes that initially formed continu-ous prototypes with the correctly identified corn.

Feature Mapping involves a number of mutuallyexclusive cell identifications. I have already men-tioned that once a feature is marked and assigned toa feature class, it is excluded from consideration forother feature classes. Similarly, cells that are proto-type features or marked features cannot becomeprotected areas unless their feature status is firstremoved. You can remove prototype status using

the drawing tools with Draw Pro-totypes as the selected action(choose Exclude instead of In-clude with the Select Area toolor hold the key with the

Select Point tool), but for this example wewill use the special function that removesthe prototype status of all unmarked fea-tures.

Note that when the prototypes are cleared,the marked feature remains, and the samplerange values are unchanged even thoughthe sample cells are now part of the markedfeature.

When you use Clear Proto-types to remedy a problem,you need to take some otheraction, such as removingsome sample points or pro-tecting areas, or you will getthe same prototypes the nexttime you click on the Classifyicon.

Feature Mapping

protectedstrips

Protecting Areas to Separate Classes

click on the DefineSamples iconclick on the Classifyiconadd more samplepoints, then click onthe Classify iconand repeat until thecorn is well definedclick on the MarkFeatures iconclick on the corn (proto-type between the twoprotected strips)

marked feature forneighboring cells, youneed to provide a pro-tective strip at theboundary between celltypes so the marked fea-ture does not spill be-yond its actualboundaries. This prob-lem is generally notapparent until after theclassification step, soyou need to use theClear Prototypes func-

STEPSclick on the ProtectAreas iconat 6X zoom (as on page18), locate the pivots be-low and just to the right ofthe marked featureusing the Select Pointtool, draw two protectivestrips as shown (over firstcolumn of lighter cells inthe upper right pivot andat first column in the leftpivot next to the boundarybetween the lower pair)

Remember that Feature Mapping is an interactiveprocess that keeps you, the “expert,” in the loop tomake decisions about the validity of computer iden-tified prototype features. Protected areas are ex-cluded from consideration in subsequentclassification steps.

The misclassification of cells that can be preventedby protecting areas can occur in two ways. In somecases, as in the current example, cells adjacent to thefeature class of interest are also identified as proto-types by the sample range defined. To prevent theseadjacent cells from becoming part of the correctly

tion before proceeding. In other cases, areas requir-ing protection are not identified as prototypes butare completely surrounded by prototype cells. Theseareas will become part of the marked feature if holefilling is on. Such areas can be protected withouthaving to clear prototypes.

corn

Feature Mapping

Drawing Prototypes and FeaturesSTEPS

scroll if necessary to lo-cate the next set of pivotsshownclick on the DrawPrototypes icon

The sample points added to mark the second featurewere sufficient to almost completely identify addi-tional pivots planted with corn. Some nearby fea-tures can be easily filled in by drawing a fewadditional prototype cells.

last marked feature

use the Select Pointtool to make a straightright edge and connectcells at the top of the pivotturn on Hole Filling if it isnot already onclick on the MarkFeatures iconclick on the newly modi-fied prototype feature andon the two connectedpivots at the right edge ofthe imageclick on the DrawMarked Featuresicon (tool changesto Select Area)draw a polygon that in-cludes the dark red cellsnot yet marked around thetwo pivots at the edgeright-click or click on theInclude button

The sample range re-mains unchanged whenyou draw either proto-types or marked fea-tures, so generally youwould repeat the add-ing sample points / clas-sification cycle to see ifthe entire feature couldbe classified as a proto-type. Sometimes add-ing more samples andclassifying again intro-duces too many proto-

types to proceed with that strategy.The point is, there are many toolsavailable in Feature Mapping, and

you can mixand match toachieve the de-sired results.

Feature Mapping

Finish Marking Irrigated CornAt this point you need to use a combination of thetechniques already learned to identify and mark therest of the corn pivots, which are identified for youin the vector map of all the pivots in this MSS scenebelow. I would recommend starting with the pivotsclosest to those already marked to define more samplecells and classify iteratively until some prototypefields are well defined. Use the drawing tools toexclude some cells from the prototypes if necessary,then mark these features. You can also draw markedfeatures using either the point or polygon tool. Youmay need to remove sample points at the extremes ofthe range and clear prototypes if a classification stepproduces results that overrun adjacent fields. I rec-ommend adding only a few sample cells betweenclassification steps so that it is not difficult to iden-tify and remove offending sample points. You canalso introduce more protected areas if need be.

Once you have features (or categories) saved as araster, you canget measurements of individual fea-tures using the GeoToolbox. Simply position thePoint tool on a feature, right click and choose theRaster Solid Trace region tool from the menu, selectthe features raster in the Region Generation windowand click OK. The Measurement panel in the Geo-Toolbox provides feature measurements, extents,and the centroid location with the region tool se-lected.

STEPSclick on DefineSamples and expandthe sample set for cornclick on Classifyrepeat the first twosteps as necessaryremove extreme samplepoints and clear proto-types if the classificationresults are too messyclick on the DrawPrototypes icon andadd to or subtract fromthe prototypes generatedas necessaryclick on the MarkFeatures icon andmark corn fieldsclick on the DrawMarked Featuresicon and edit markedfeatures as necessarysave your features rasterobject (File / Features /Save)click on theGeoToolbox, choose thePoint tool, left-click on afeature, then right-clickand choose RasterSolid-Traceclick on [Raster], selectyour saved features,and click [OK]

marked corn

MSS scene vector pivots

Feature Mapping

aren’t clear on which piv-ots and partial pivots inthe view are soybeans,add more soybeansample points as neces-sary or draw featuresmark the soybeanpivotssave your marked fea-tures and style object

More than One Feature ClassBefore you begin to work on identifying anotherfeature class, you need to clear both the samples andprototypes. Or, if you exited after completing thelast page, just open the features raster (File / Features/ Open); there are no sample points, prototypes, orprotected areas to clear. You also need to establisha new feature class sometime before you mark thefirst new feature.

When there is overlap in cell values for differentfeature classes, as there is in this example, you’ll findit gets progressively easier with each subsequentclass to identify prototypes and mark classes be-cause cells from the first feature type marked are no

select Clear Samplesand Prototypes thenUnprotect All from theSpecial menurepeat steps 2–4 onpage 18, except this timedefine samples inthe left half of thepivotclick on the Classifyicon; add moresamples as neededrepeat the steps on page8 to add a feature classand name it soybeans

STEPS

longer under consid-eration. Now that thecorn is marked, you’llfind it is easier to iden-tify and cleanly markthe soybeans. Whendefining samples, pre-viously marked fea-tures are displayed intheir assigned color soyou know they are in-eligible for samplepoint selection.

When you have mul-

consult the cropmap (right) if you

othersoybeanscorn

tiple feature classes in theFeature Class Selectionwindow, you need tomake sure that the featureclass you want is selected(black outline) when youmark/unmark a feature.

Feature Mapping

Creating CategoriesCategories are used to group features in the reportgenerated by the Feature Mapping process. Thus,you can look at the number and area of features ofspecific types by land ownership or any other at-tribute you choose. There are two components toany category: the ground area to which the categoryrefers and the category’s style, which includes itsname and assigned color. We will create hypotheti-cal land ownership categories to use in generatinga report that includes the acres of corn and soybeansplanted for each land owner in the area.

The mechanics for creating categories are much thesame as for drawing marked features. Once an areais assigned to one category class, it cannot beincluded in a second class. Thus, you need to becareful only when drawing the first category classboundaries; adjacent categories will share thisboundary regardless of how they are drawn. It isactually a good idea to draw later categories so theirboundaries overlap previously assigned areas sothere are no unintentional gaps between categories.Categories are only shown in the Feature Mappingwindow when Categories is the action selected onthe toolbar.

click on theCategories iconrepeat the steps onpage 8 for adding fea-ture classes five timesto create five categoryclassesclick on the first classname you are going todrawdraw internal edges of theboundary polygon care-fully and extend outeredges beyond edge ofrasterclick on [Include]click on the name of thenext classcarefully draw the edgesthat don’t bound the firstcategory or the edge ofthe raster; draw otheredges so that they extendinto the first category orbeyond the edge of thereference imagerepeat the last two stepsuntil all areas have beenassigned to a categoryclass as shownsave your categoriesraster and style objects(File menus in the viewwindow and CategoryClass Selection window,

respectively)

Feature Mapping

area(acres)

Generating a Report with CategoriesCategories are reflected in the Feature Mappingreport in three different ways. A general summaryabout each category is included (Record Type CS).This information includes the category name, thetotal area covered by the category, the number ofcells in the category, and the percentage of the totalarea that belongs to the category. Each feature is alsobroken down by category (Record Type FB). If anindividual feature is entirely within one category,this information will be the same as the featuresummary record (Record Type FS). Feature totals foreach category are also provided (Record Type CT).In this example, there may be three feature totalrecords for each category: corn, soybeans, and un-classified.

follow Steps 2 through 6on page 14 to generateand view a Feature Map-ping Report

CS Johnson 1850.66 2307 5.2CS Michaels 6809.01 8488 19.2CS Chase 16117.66 20092 45.5CS Granger 4178.62 5209 11.8CS Taylor 6472.09 8068 18.3

Marked corn andsoybean featuresoverlaid on cat-egories raster.

FS soybeans 7 3 51 68.99 7.01 86 123FB Johnson 7 3 51 68.99 7.01 86 123 soybeansFS corn 8 175 68 215.79 19.67 269 345FB Michaels 8 175 68 4.01 0.46 5 8 cornFB Chase 8 175 68 211.78 19.22 264 337 corn

Feature that crossesMichaels and Chaseboundary

Feature entirely onJohnson land

feature class (FS) orcategory class (FB)

featurenumber

centroidline & column

boundarylength

featureclass (if

FB)

area(acres)

categoryclass

numberof cells

percentageof area

cells inarea & boundary

CT Johnson Unclassified 1 1781.67 0.00 1781.67 0.00 96.3 2221 0CT Johnson soybeans 1 68.99 7.01 68.99 7.01 3.7 86 123CT Michaels Unclassified 1 6241.86 0.00 6241.86 0.00 91.7 7781 0CT Michaels corn 4 206.16 21.95 51.54 5.49 3.0 257 385CT Michaels soybeans 4 360.99 36.21 90.25 9.05 5.3 450 635CT Chase Unclassified 1 14299.89 0.00 14299.89 0.00 88.7 17826 0CT Chase corn 8 622.50 61.81 77.81 7.73 3.9 776 1084CT Chase soybeans 9 1195.27 119.29 132.81 13.25 7.4 1490 2092CT Granger Unclassified 1 4002.94 0.00 4002.94 0.00 95.8 4990 0CT Granger corn 1 15.24 1.94 15.24 1.94 0.4 19 34CT Granger soybeans 4 160.44 17.79 40.11 4.45 3.8 200 312CT Taylor Unclassified 1 5369.88 0.00 5369.88 0.00 83.0 6694 0CT Taylor corn 4 620.09 58.91 155.02 14.73 9.6 773 1033CT Taylor soybeans 5 482.12 47.33 96.42 9.47 7.4 601 830

categoryclass

featureclass

numberof features

area(acres)

boundarylength

averagearea

cells inarea & boundary

avg. bound.length

percentof area

Feature Mapping

AfterwordYou have created two simple feature maps and usedmost of the tools available. Using Feature Mappingwith your own imagery may require more than twoclasses. The steps are the same regardless of thenumber of classes you want to map. Just remember,the class you are marking or unmarking needs to beselected in the Feature Class Selection window. Ifyou mark a feature as belonging to the wrong class,you can simply unmark it and remark it with thecorrect class selected unless it happens to merge witha previously marked feature. If you accidently marksoybeans, for example, as corn and it adjoins analready marked corn field, you have a single, largercorn feature. Because you cannot unmark part of afeature, correcting this error will likely involve re-drawing the corn feature after unmarking then reclas-sifying and marking the soybeans.

One simple idea not introduced is the decision rulesettings (Options menu) for classification. The de-fault Rule setting uses a simple boxcar classifier thatestablishes the area of n-dimensional space (where nis the number of rasters selected for analysis) whereprototypes are identified by the range of samplevalues. Quantitative data is required for such anapproach to provide meaningful results. Individualgrayscale bands or 16- or 24-bit composite color areappropriate data for the default Rule setting (Range);8-bit composite color is not. If 8-bit composite coloris all you have available for analysis, you need to usethe Exact decision rule. Prototypes identified by theExact decision rule are only those cells that matchthe set of cell values defined by one of the samplepoints.

An important concept that was only mentioned inpassing in this booklet is the simplicity of creatingvector polygons from your features rasters. You justselect the features raster and run the Auto-Bound-aries process and you have vector polygons with thecorrect feature class assigned.

Run Auto-Boundaries to gen-erate vector objects fromfeature maps (Process /Convert / Raster-to-Vector /Auto-Boundaries).

CIR

TVI

PCA

Advanced Software for Geospatial Analysis

M icro Im ages, In c .

FEATURE

MAPPING

Voice: (402)477-9554FAX: (402)477-9559

email: [email protected]: www.microimages.com

11th Floor – Sharp Tower206 South 13th StreetLincoln, Nebraska 68508-2010 USA

MicroImages, Inc. publishes a complete line of professional software for advanced geospatial datavisualization, analysis, and publishing. Contact us or visit our web site for detailed productinformation.

TNTmips TNTmips is a professional system for fully integrated GIS, image analysis,CAD, TIN, desktop cartography, and geospatial database management.

TNTedit TNTedit provides interactive tools to create, georeference, and edit vector, image,CAD, TIN, and relational database project materials in a wide variety of formats.

TNTview TNTview has the same powerful display features as TNTmips and is perfect forthose who do not need the technical processing and preparation features of TNTmips.

TNTatlas TNTatlas lets you publish and distribute your spatial project materials on CD-ROM at low cost. TNTatlas CDs can be used on any popular computing platform.

TNTserver TNTserver lets you publish TNTatlases on the Internet or on your intranet.Navigate through geodata atlases with your web browser and the TNTclient Java applet.

TNTlite TNTlite is a free version of TNTmips for students and professionals with smallprojects. You can download TNTlite from MicroImages’ web site, or you can orderTNTlite on CD-ROM.

Indexadding feature classes ....................... 8, 24analysis rasters ................................... 4, 17categories .......................................... 25, 26Classify function ................... 7, 18, 21, 24clearing prototypes ................................. 20colors ....................................................... 1 8decision rules .......................................... 27defining samples ................... 6, 18, 21, 24dimensional reduction ........................... 17drawing prototypes and features .......... 22exclude / include features ...................... 12feature classes ............................................ 8feature mapping report ................... 14, 26Feature Mapping toolbar ......................... 4hole filling ............................................... 16identifying prototypes ............................. 7include / exclude features ...................... 12LegendView .............................................. 7marked feature, definition ....................... 5marking features ..... 9, 11, 13, 19, 21, 23

measuring features ................................. 23Principal Components ............................ 1 7protecting areas ....................................... 21prototype features ......................... 5, 7, 11reference rasters ................................. 4, 17region of interest .................................... 15removing sample cells ............................. 6report ................................................ 14, 26restarting feature mapping .................... 11sample cell range ...................................... 6sample points ..................................... 5, 11saving features ........................................ 1 0Select Area tool ......................... 12, 13, 22Show Values option ...................... 5, 6, 18style object ................................................ 8Transformed Vegetation Index ............ 17unmarking features ................................ 12vector polygons from features ............. 2 7View window ............................................ 5vignetting .................................................. 3

Front CoverBefore Getting StartedIntroductionAnalysis & Reference RastersFeature Mapping View WindowDefining Sample CellsIdentifying Prototype FeaturesAdding a Feature ClassMarking All FeaturesSaving ResultsResuming the ProcessUnmarking FeaturesMarking Features with the MouseGenerating a ReportDefining a Region of InterestFilling HolesMore Complex Feature MappingDefine Samples and ClassifyDealing with Messy PrototypesClearing PrototypesProtecting Areas to Separate ClassesDrawing Prototypes & FeaturesFinish Marking Irrigated CornMore than One Feature ClassCreating CategoriesGenerating a Report with CategoriesAfterwordIndex and Back Cover

e f tutorial atur ap feature pin mapping · 2003. 10. 28. · space used for identification of...

Documents