PIXEL BASED AND OBJECT BASED CLASSIFICATION FOR SHORELINE DETECTION Prepared by : Nita Maulia Bayu Andrianto Wirawan “4s” Magister for Planning and Management of Coastal Area and Watershed . Geography Faculty Gadjah Mada University ~ 2010Introduction + Why Shoreline extraction? Shorelines are recognized by the International Geographic Data Committee (IGDC) as one of the 27 most important features to be mapped and monitored (Li et a/., 2002 in Nguyen et al., 2008) * Why using Landsat? Landsat can provide comprehensive images + Why Pixel Based and Object Based? Pixel based methods are strongly believes as a powerful method to extract information, while object based methods that categorized as a new methods, is possible to be applied in shoreline extractionOutline This guideline book describes the workflow of shoreline detection using the Density Slice (Pixel Based Classification) and Region Growing Segmentation (Object Based Classification). Envi, Spring, and GIS software were needed to run the shoreline extraction process. Make sure that those software were already installed on your computer. The data and the location that used in this exercise were part of thesis project “Shoreline Change Analysis and Prediction; an Application of Remote Sensing and GIS” by Maulia (2010).Pixel Based Classification Getting Started... Basic Tools Classfcation Load the image by select Fle Options BP 1992 image end 1 File > Open Image File > Open Vector Fie choose the directory where Cane? the file was saved. The 2 @ Mopinto Open External Fle » Available Band List Open Previous File > windows will pop-up GrayScale AGB Color Selected Band Density slice only could be done to the single [BendS1952megeuh band, and the best band to determine the water bodies is band 5. To display the image, click on the desired band and then choose the Gray Scale radio button. Load the image by click Load Band in the New Display Dine] 720 x 1070 Exe) 50] Lond Band | Ne Dissay =|Density Slice Classification In the display window image select Overlay > Density Slice. Select the input band (Band 5) for Density Slice in the pop-up window then click OK. Density Slice parameter will be set in the next pop-up window. The pixel values were automatically classified by the programs into some range. Since this method will be applied to extract shoreline, the pixels will be classified into two ranges that present the water/sea and the land. Change the Min and Max value into 0 and 255. Select Clear Range below the Density Slice Range box to remove the ranges. To create new ranges select Option > Add New Ranges. Specify the ranges arbitrarily based on the image histogram, for example 0 — 15, 16 — 255, etc, then Apply to the image.= Select ng Dens Shoe Band font SP 82 mae wore © Band! Bard? file optine He re re ata Bods — [Ppastao Foe at chao bey Ske and Banos [HE Mal 82 et gona Ranges © @ Mop no Sette of laa Panges.. Daina Deal Stes Fangs ed Ste dsteem) i Feicsrh se rosacn =e 2h 144 agent Range Star foe ace Se Eat Pergo | Dstt Renae] Ce Raroe J] ‘09 || werdow FF tnage/Zoam 1% Sect A584 |] v4) Staning Cotoe FT] otRanges[t Sh DataRarge, eset = | Clas Rapes [eoeeT redone tnageaoon SexThe output of density slice need to be convert into vector file. Select File > Output Range to EVF-s then selects the Ranges and save in the directory where the data will reside. Available Vector Layer window will appear. Select the vector layer then Load Selected vector into New Vector Window. In the display vector window, select File > Export Active Layer to Shapefile to convert the file into shapefile. Specify the directory to save the file. Edit the classification result shapefile into Shoreline feature using edit tools on your GIS software.Cpe Heb Sve Rare | Restore Rares em 3 Costu Rages tC nape cancel Selec Ranges EVES 5 Pa Crlsfed Co : Sas a= eaten A hiatal te elect [2 ceca ee A006] 4 seca ten | Ces At as ups Select Vector Wid EW Lave: Desiy Sle Fees [1882 rae) ipl = OuputReatio © Fle |__| —__—__—- Rave Deny Senge [ame Oar Dc, > cna tuparenmeton {ee CUseae rd SeeLDoanet « rte: Oup Fenane | se 6224 oe re |Recerds: 1,505 [23,858 nodee) EXDosanets and Senge 2 Ene Dap ae Dose ok | ConclGOOD JOB Now, answer these following questions to be included on your report. 1. What is the analysis input used in Density Slicing process? 2. Explain the classification result regarding to the condition of the area and your arbitrarily ranges. 3. Is the water and land can be separated properly? please explain. 4. In this case, could the shoreline be extracted? Please explain.Object Based Classification Create a New Project Before start the image processing, new project directory and project need to be created. To create the project directory, select File Database. New window which show the requirement for creating a new project will pop-up. Specify the requirement; directory location and name then click Create > Apply > Close. [ae Pree arama) [xen a] eat | Seay esr bean Inger Ti] Inga Inger Pe] Inger TF ‘Specity the directory name| [ine oneal pe na = conn: te (a) nage Repsaton Sprit Teal conve, Ba (al ft](cst view % Databes OG fe Data Model. New Shortout Inport. Load Shortcut. = Import Table. Inport from Sar. Inport Redistrated images.. : | vert to ASCII Select File > Project to create Saag a project. When pop-up window was = appeared, specify the requirement Export SpringWet box. Specify the project Name = (e.g. Project1), Projection, and Bounding Box as _ displayed below. Ae JPEG I Image Registration. Spring to Terral converter... ExitTo set the projection click at the projection box and then specify the Projections requirement as follow; System (UTM) > Earth Models (WGS84) > Long (e 111 00.00) > Zone (49) > Apply > Close. ‘Specity the project name [rin] utube Terese hiioomo Contests © Geapoe @ Hones Staind Pre roniphee: ON @ 5 nes Cc =, u ‘iia [Eanes — Te At the bounding box, choose Planes option for Coordinates. Filled X1, Y1, X2, and Y2 with the following value 438189.75, 9230402.0, 459789.75, and 9262502.0. Don’t forget to choose the Southern Hemisphere at the end of the options. Click Create > Load after finish filling the specification.Adding the Data to the Project Project that created before was not contained data yet. To start adding data to the project select File > Data Model, when the pop-up window appear create a category by specify the category name in the Name box (e.g. Landsat_1992) and then choose Image for the category Models. Click Create > Apply > Close. (Open Image CBER: Image Registrat Database... Project Inport. Inport Table... Inport From $6 [Specity the categories Inport Registrate Tine Canes ae Convert to ASCH Inport TIFF/GeoTh Export. Export SpringWeb] Export to Terratb} Save As JPEG Ima} Spring to Teralo Ext Ton © image Thema object [_ceae [lick image in the lode button change network, Nonspatisfe a a — Database = foe Fin) eiDocaners ad ste [a Select File Import esi Some TIFF/GeoTIFF then load the Males ; a File from the directory, set the Inpet from Sa Barer on Dummy Value for 255, ere ere choose the Entity as Image, ae then load the Category that Qpeninege CBERS. sone created before and specify the Ext H - Infolayer (e.g. Band). meena lt Save Be FE nage. Click Apply > Close. The : = data will loaded to the display ae window Sprig to Terral convert EatTo display the composite image in the window, select the Band infolayer and chose the button bellow at which the band will be displayed. For example Band 2 will be display as Blue band, select the band and then click at the B option below the categories option. Pearse Petr Teas hab LAAGSN BODY (avo )fniees ose) @ Sse s+ Select the one of [+ the infolayer bas Select one of the optional bands todisplay aea|es = = an [Ca Cane Cae CaSegmentation Cont. Feeteg To start segmentation process, in the tool bar — sthwercoperetoes. Fafa FRegon Growing [= select Image > Segmentation. In the M<*#tson appearing box define the Method that will be used in segmentation process as Region 3 [lands 1992) Bond. e Plardeat 1992] Band. 2 Z [landeat 1902] Band . Pra Compan Growing. espa Casesfra1” fit |st_ signe) 00 Select all the image's bands in the Bands option —“xns#comner E and define the Similarity and the Area oma. Eee Bae parameters, for example use the parameter of setininsion | similarity 11 and Area 2000. ae a3 . Since all the bands are included in the joc ssemacors cut segmentation process, choose none for Band — sxt->cundrane [eotedimage: [Smit arcood) Exclusion. Specify the output name for the — raswesns wcSmoctting: _@ Yes_© No segmentation result in the Segmented Image = «7k ounding Box) (e.g. Sim11_Ar2000) and click Yes for Arc SS oe) Smoothing radio button. (acim) (nia! Click Apply >Close to start the process and after the processing finished the dialog box ta i ee below will appear, click OK. The segmentation result will be displayed in assistant display window eal eriegn 03/5 & 3 S2n #5 End Day Sat chsh 475Classification The segmentation result needs to be classified. To classify the segmentation result, select Image >Classification. First step that need to be done when the classification window popped-up is create the classification Context Creation which can be accessed by clicking Create in the Context box. contrast Feng. Aetheti Operations ee Esoonewasen Pomerat TAS<->AGE Tanfermatin, Pinal onponents ture Model Seatetato, Eom ageing Classes fra Thema Corrected ConponeetsLaball sande Sates, Restoran. Nase Elation Pred Vales, Pater Antenna Corecien, Slane -> Ground Range Con aster Edna Syeti >A Specify the Name of context (e.g. Class). Cy Anas Type_O) el fans [Laat 1950] ard 9 [eats -bard¢ lara 195] -Bare 5 [eat 1950 “Bard 6 [area tna Choose the Regions buttons for Analysis Type, included all the Bands for the classification by clicking them, and the last choose the Segmented Image as the classification input. Click Apply Close.Select the Context that was created before and then click at Extraction of attributes of the regions so that the training area for the classification can be made. Training area can be made by select the Training box and then the dialog box that command the sample was acquired will appear. Click OK and start to create training area Treen) FlssonearonairoaT = eat) [oe ‘rd ard 2 Segmeted nap: [mle aaanet aban d tevepr] anes eaten] Post Clasafcaon, ‘passin, J i)Training Firstly define the Name of the theme that will be create (e.g. Water and Land), the Color of each theme also can be defined, then click Create. The theme will be added to the theme box. Select one of the theme on the theme box, choose Normal radio button for Mode and Acquisition radio button for Type. Click to the segmented polygon in the image as the_sample for the theme by using |. Once it was selected, click Get to take the segmented polygon as a sample. Take some sample for each theme Nave: [nd Thane aed ede: @ Nxmal © Grup © Ungo.p [show at Type: @ Acguston Test foyaen ©) Rectangle _@) Reson (seas) at Neon Neral Fl (105) Feber Edtrg Thames pert ml feaiatin [| (an oma) GaneTo save the classified image, new category needs to be created. Create, 4, new category by selecting File > Data ene! Model and then create a category Name. Choose the Thematic on ro Models and then create Thematic iment Table Classes by specify the class Name and then click Create (e.g. Land and Water) . Clik Thematic in the ‘Made radio button Project Export Themes by specifying the Category. Choose the Thematic 1 model categories (e.g. Map). Type the : econo infolayer (IL) box with a name (e.g. Sim11_Ar2000) and make sure that the Acquisition is appearing in the box below IL. Click Export to export == the theme in the display window. cere After finish it, click Save >Close.To start the classification, click Classification on the Classification box. Create the classification by specifying the Name (e.g. Classification) and then click Crate. Choose Bhattacharya as the Classifier with the Acceptance Threshold 99.9%. Click the Sample Analysis to check the accuracy assessment. Click Classify to process the classification. Namar [classfication create Classifier = Thewer Theme Confuron Mati and tend 0.09% Rejection "0:00% [Eareie anatase] enContrast. Fiteng ‘rthmetic Operations, THS<->RG8 Tranefomation. Frinepal Conpanert, sture Model Segmentation. section, in Connected Components Label, Statist, Restoration. Noe Elination, Pel Values Mosaic Pattern antenna Correction Sant -> Ground Range Conversion. Raster Edtng Synthetic > RGB Assistant window will appear and in the Map es The classification result is in raster format, meanwhile the vector data is needed for further analysis. To convert the classification result into vector data, click on the Classification result in infolayer, then select Image > Mapping Classes for a Thematic Image. The dialog box will appear and then specify the Theme according to the Classes and click Apply after finish. fon Vector Edting. Raster Eating... Mosaic. Label Placement. Themes Mlccces Measurements cf Clas Coss Tabulation, Active Layer: [ClasifiioneT Tre Snocting@) Yes Conversion Type: _ foal : categories new infolayer was added as Classification-T. Select the new infolayer, display it, and then in the tool bar select Thematic > Raster -> Vector. (al Ga eslExport the result by selecting File > Export then dialog box will prompt. Choose SHAPEFILE as the format, and the Planes(meters) as the Coord. Click Save and specify the directory where the file will reside and the file name and then click Close. The Polygon and Polyline shapefiles from the classification will automatically add to the directory you have specified before. Once the shapefile is resulted, the next process to produce shoreline will be much easier. The standard edit tools in your GIS software were can be used to do this process. Database, Project Data Node Inport Inport Tale Cond: Pares(tes) Sc hes Spring to Tera converte, ExtWOW!! ANOTHER GOOD JOB!! Now, there are another questions that have to be answered to enrich your report. 1. What is the analysis input used in Density Slicing process? and what does Similarity and Area parameter means? . Explain the classification result regarding to the condition of the area and your arbitrarily parameters values. . Is the water and land can be separated properly using RGS method? please explain. . In this case, could the shoreline be extracted? Please explain. . Compared the DS and RGS result. Are they shows any differences? . What are the advantages and disadvantages of using DS and RGS for shoreline extraction? . Which one is the best method for shoreline extraction? Please explain.