Georectification and Image Analysis Techniques using ArcGIS 10

Prepared by Parikshit Ranade, and Ayse Irmak

GIS in Water Resources

Fall 2011


 
Purpose

 

The purpose of this exercise is to perform typical tasks of remote sensing analysis using ArcGIS 10. We will perform georeferencing, and NDVI computation. We will use Google captured image as an example for georeferencing. With this exercise you will –

 

1. Capture an image from Google for Lincoln, NE and georeference it.

2. Compute a Normalized Difference Vegetation Index (NDVI) of Landsat image

 

Computer and Data Requirements

To carry out this exercise, you need to have a computer, which runs the ArcInfo version of ArcGIS 10. There is no need for remote sensing software. The data are provided in the accompanying zip file.

 

Downloading Landsat data

Landsat data can be downloaded for free from http://edcsns17.cr.usgs.gov/EarthExplorer/ or http://glovis.usgs.gov/. USGS global visualization viewer facilitates viewing and downloading of satellite imagery. USGS provides Landsat 5 TM data in TIFF format. This data can be converted into other formats using ArcGIS or any image processing software (i.e. ERDAS-IMAGINE, ENVI). Landsat satellite passes over earth, the area can be identified by the path and row combination. Our study area is under the path 28 row 32. This is southeastern part of Nebraska and Landsat passes over this area at approximately 10:15 CST. Each Landsat scene has a unique identifier. Our dataset has identifier LT50290322005251EDC00. All the information regarding Landsat data can be found in Header file of Landsat data.  These files will end with _GCP and _MTL. These files also contain the information regarding each band pertinent to that scene. Useful information like center and corner coordinates of Landsat scene is also provided in the header file. A screen shot of sample attributes for this Landsat scene is provided below.

 

Landsat 5 TM has 7 different bands. These bands are useful for extracting various information related to vegetation, temperature, clouds, soil moisture, biomass, rocks, minerals and so on.  Below are details of band designation for Landsat 5 TM. More detailed information can be found at http://egsc.usgs.gov/isb/pubs/factsheets/fs02303.html.

 

Band

Spectral Bands1

Wavelength  (micrometers)

Potential Information Content

Resolution (meters)

Band 1

  Blue

0.45 - 0.52

Discriminates soil and rock surfaces from vegetation. Provides increased penetration of water bodies

30

Band 2

  Green

0.52 - 0.60

Useful for assessing plant vigor

30

Band 3

  Red

0.63 - 0.69

Discriminates vegetation slopes

30

Band 4

  Near IR

0.76 - 0.90

Biomass content and shorelines

30

Band 5

  Mid  IR

1.55 - 1.75

Discriminates moisture content of soil and vegetation; penetrates thin clouds.

30

Band 6

  Thermal IR

10.40 - 12.50

thermal mapping and estimated soil moisture

120

Band 7

  Mid  IR

2.08 - 2.35

Mapping hydrothermally altered rocks associated with mineral deposits

30

1: IR stands for infrared

 

Various composites can be mapped to facilitate viewing of the raw imagery. A false color composite of near infrared is useful for observing spatial distribution of the vegetation in the Landsat scene by combining bands 2, 3 & 4 (RGB). We are going to generate the true color composite by using spectral bands 2 (green), 3 (red), and 4 (near infrared).

 

Part I – Georeferencing

 

Raster data or remote sensing imagery served by various state and federal agencies usually comes with  spatial reference. However, sometimes the location information delivered with them is inadequate and the data does not align properly with other data layers you may have. Furthermore, there are cases that raster data was digitized using scanned paper maps; thus, the co-ordinate system needs to be assigned.  Therefore, to use a raster datasets (i.e. remote sensing imagery) in conjunction with your other spatial data layers, you may need to align/geo-reference your raster datasets. Georeferencing a raster allows the raster to be viewed, queried, and analyzed with other geographic data.

 

In this exercise, we will use Google map as a data source to download an image which needs to be georeferenced with other data layers in the exercise. We can use Bing maps directly in ArcMap v10 as  a basemap. However, we are going to use Google map in order to illustrate “how we can add spatial reference to our own map/image”. Your data could be a “scanned” paper map of your field data or any other scanned hard copy map (soil survey, groundwater level, nitrate concentration etc.)

 

Follow step by step procedure below to georeference the image.

1.      First go to http://maps.google.com/ on your web browser. Search for Lincoln, NE. Click on the Satellite option on upper right corner of Google map window. On the Northwest side of the map you will see Capitol Beach Lake near I-80. The Interstate 80 is a major transcontinental corridor connecting California and New York City in the USA.

 

 

2.      Now click on the options button  on the upper right corner of the webpage.  Click on the Maps Labs option. You need to enableLatLng Marker” tool.  The toolwill allow you to mark the longitude and latitudes of the locations of the map. Optionally, you can enable “LatLng Tooptip”. 

 

 

 

3.      On the map, Right click on any point for which you need to mark the latitude and longitude. Select “drop latlng marker” option. To zoom in closer to the point, right click and select “zoom in” option. It is very easy to use this compared to “zoom in bar” on the Google (on the left).

 

You could georeference a raster using existing spatial data (target data), such as a vector feature class, that resides in the desired map coordinate system. The process involves identifying a series of ground control points with known x,y coordinates. This will link ground control locations on the raster dataset with locations in the spatially referenced data (target data).

 

Control points (ground truth) are locations that can be accurately identified on the raster dataset. There are many different types of features that can be used as Control points, such as road or stream intersections, the mouth of a stream, rocky outcrops, the end of a jetty of land, the corner of an established field, street corners, or the intersection of two hedgerows. 

 

4.      Similar to shown in figure on the next page, try to identify the control points which are unique and easy to locate on the map.  Avoid control points which are too close to each other. Also try to find control points from different locations on the image.

 

5.      Capture the screen using  PrintScreen button on keyboard and paste it into Paint program and save it as a TIFF image. Paint program can be opened on windows PC from Start/Programs/Accessories/Paint. Note down the Latitude and longitude information for each control point and save the image as ControlPts.TIF. This will be used as your reference if you enter the control points incorrectly. Therefore, this image should include location and coordinates of the “control point” that you will utilize to georeference the “capitol beach” image. 

 

 

 

6.      Get back to your google map browser window. Close the “LatLng Tooltip” in the google map by clicking on the  button.

 

7.      Uncheck the “labels” option under “Map” option in the upper right corner of Google map window (figure not shown).

 

8.      Again capture the screen, open the paint program and paste the image. Crop the image to clean the sides of the image. After cropping, the image should look similar to image on the next page.

 

9.      Now “Save” the image with a name “Capitol BeachRaw.TIF under Ex6 folder. We cannot save it in geodatabase directly from Paint program. We will save it in Geodatabase later.

 

10.  Now open the ArcCatalog and under the Ex 6 folder right click to create new/personal geodatabase. Name it as EX6. 

 

 

11.  Open the Capitol BeachRaw.TIF  image in ArcMap. Right click on Capitol BeachRaw.TIF, click on the Data/Export data. Select name as Capitol Beach and location as EX6 personal geodatabase. Add the Capitol Beach.TIF to ArcMap.

 

 

Now you will see clear satellite image of Capitol Beach without any labels or location information in the ArcMap.

 

 

We will open the ArcMap and set the coordinate system of the dataframe for proper alignment of the layers displayed in ArcMap.

 

12.  Right click on layers on upper corner of Table of Contents. Go to Properties\coordinate system/import. Select the Lancaster_County polygon in Lancaster geodatabase. This will change the coordinate system of the data frame to NAD 1983 StatePlane Nebraska FIPS 2600 Feet

 

 

Now we have identified the control points (ground truth) with Latitude and Longitude information, we will import them in ArcGIS.

13.  Open ArcCatalog. Right click on Ex6 geogatabase, select new/feature class. Name it as Control_Points alias Control_Points. Select point feature type.

 

14.  Select the Spatial reference in Geographic Coordinate System/World/WGS 1984.prj. Leave rest of the options as default. This is the coordinate system that Google uses.

 

 

15.  Now lets start adding each control point to the feature class. Go to Toolbars/Editor to open editor toolbar.

 

 

16.  Click on Start Editing to edit the Control_Points layer.

 

17.  Open the Capitol Beach.TIF that you have saved under Ex6 geodatabse.  You might receive warning about unknown spatial reference. Click OK to ignore the warning.

 

 

18.  In create feature window on the right side of ArcMap choose create control points layer. In the same window under construction tools choose point.

 

19.  Right click on any point within layer in the ArcMap window and select Absolute X,Y

 

20.  Type in X (lat) and Y (Long) value that you have recorded from Google LatLng marker for each control point (ground truth).  Make sure you assign negative (-) sign for the longitude of the control points (ground truth). Make sure X, Y unit is Decimal degrees.

 

 

 

21.  Repeat the same procedure to add all the control points (ground truth) that you have selected. If you enter the wrong LatLong information for the control point, you can delete the point. Click on the control point to select it and press delete key on the keyboard to remove it. You can also right click on the selected control point and click delete.

 

22.  Click on the Stop editing button on editor toolbar and click yes to save your edits. Now you have completed the control points feature layer. This was to illustrate one way of adding control points.

 

Another simpler way is to add control points using Make XY event layer tools from text or excel table (steps 23 through 28).

23.  Open MS Excel program. Create CPID, Lat and Long columns in Excel. In the excel table enter the co-ordinates of control points in Lat and Long column. CPID is your control point ID for your reference.

 

24.  Remove the remaining worksheets from the excel. Save the Excel file with name CP.

 

 

25.  In ArcToolbox, go to data management\layers and table views\ make XY event layer. 

 

26.  To select XY table, browse to CP.xlsx file which you have just created. Note you will see two files within CP.Xlasx namely _xlnm#Database and CP$. Choose CP$ and click add.

 

27.  Choose X field as Long and Y field as Lat. Leave Z field blank. Leave Layer name as CP$_Layer.

 

28.  Choose spatial reference, click select/Geographic Coordinate system/world/WGS 1984.prj. This is same coordinate system that Google uses.  Click OK.

 

CP$_Layer will be added to the ArcMap after tool is run completely.

 

 

Now we will save it as feature class in Ex6 geodatabase.

 

29.  Right click on CP$_Layer layer, click Data/Export. Save data with Name Control_Points as file and personal geodatabase feature classes.

 

30.  Note you will first have to choose ‘save as type’ as ‘File and Personal Geodatabase Feature Class’. Now you will see Ex6 geodatabase location in the ‘saving data window’. If you choose default shapefile option under ‘save as type’ you won’t be able to save it in Ex6 geodatabase.

 

Now we will start georeferencing the Capitol Beach.TIF.

 

31.  In ArcMap, go to customize/toolbars/georeferencing.

 

32.  Make sure Capitol beach layer is selected for georeferencing. Click on add control points  button.

 

This button will link our Control_Point features (ground truth) to the corresponding location on the Capitol Beach.TIF. This means we will kind of assign spatial reference to the control points (ground truth) on Capitol Beach.TIF. ArcGIS then will fit the surface and assign spatial reference to entire image. 

 

 

 

 

33.  Now let’s start adding the control points . Make sure you Auto Adjust option is on both on Georeferencing toolbar and on Link table. Link table can be opened by clicking on  button on Georeferencing toolbar.

 

You might not be able to see anything in full extent of the ArcMap. This is because Capitol Beach.TIF is missing a spatial reference.

 

34.  Right click on Capitol Beach.TIF layer click on Zoom to Layer.

 

35.  Now identify the location of one of control point (ground truth) on the Capitol Beach.TIF image. Click on the identified location on image. (Here you can refer to your ControlPts.TIF image where you have saved the Latitude and Longituide information of each control point on image.)

 

36.  Now right click on the Control_Point layer and ‘Zoom to Layer’ (your cursor will be showing line, disregard that).

 

37.  Now you will see all the control points. Click on the control point which corresponds to point on the Capitol Beach.TIF image. At this point ArcGIS snapping environment will be automatically activated and your control point will be snapped.

 

Notice your both layers are now closer. Continue adding the control points. You might receive the warning that ‘The control points are collinear or not well distributed, this will affect the warp result

 

38.  Click OK and see how much of warping (shift) is caused. If you think image is warped too much, it indicates you have misplaced the control point.

 

39.  Make sure you click on the location of control point on the Capitol Beach.TIF image first and then connect it to corresponding control point feature class (ground truth). Do not click on the control point feature class and then try to connect it to Capitol Beach.TIF image.

 

40.  If you misplace a control point, simply click on the view link table button   and delete the control point (selected control point will be in highlighted in blue as shown in figure).

 

41.  Notice once you have added the second control point you Capitol Beach.TIF and Control_Point layers will be aligned. This is due to Auto adjust function.

 

42.  Now finish connecting all the control points to corresponding location on the Capitol Beach.TIF image.

 

 

 

 

43.  Open the View link table in georeferencing window. Make sure Auto adjust option is checked.

 

44.  Compare the transformation options – 1st order polynomial (affine), 2nd order polynomial and adjust. See which image (after the transformation) looks most similar to your original image from Google. Select the best fitted transformation and click OK.

 

 

 

45.  Click on the Update georeferencing option under georeferencing toolbar.

 

46.  Click on the Rectify option on georeferencing toolbar. Change the Name of the file to ‘Captiol_Beach_Rect and Output location to Ex6 geodatabase. Leave rest of the option to default and click SAVE.

 

 

47.  Remove Control_Points layers and Capitol_Beach layer from ArcMap.

 

48.  Open the ArcCatalog. Right click on the Captiol_Beach_Rect image from Ex6 geodatabse that you have just saved. (Note: sometime you might have to reopen the ArcCatalog if it was already open, or refresh it from View/refresh or use F5 key on keyboard).

 

49.  Notice your file now has a spatial reference NAD 1983 StatePlane Nebraska FIPS 2600 Feet and datum D_North_American_1983 which is same as ArcGIS data frame.

 

50.  It is important to note that when you Update Georeferencing, coordinate system of the dataframe will be assigned to the image.

 

 

 

 

 

 Part II: NDVI computation

 

Now we will use some of the Processing functions. One of the commonly used functions to assess the presence of live green vegetation is NDVI. NDVI is normalized difference vegetation index. NDVI is computed using below formula:     RED and NIR stand for the spectral reflectance measurements acquired in the red and near-infrared regions of electromagnetic spectrum, respectively. NDVI takes the value from -1 to 1. The higher the NDVI, higher the fraction of live green vegetation present in the scene. Landsat band 4 (0.77-0.90 µm) measures the reflectance in NIR region and Band 3 (0.63-0.69 µm) measures the reflectance in Red region.  The equation ArcGIS uses to generate the output is as follows: NDVI = ((IR - R)/(IR + R)) * 100 + 100. This will result in a value range of 0–200 and fit within an 8-bit structure.

 

1.      Add ‘July’ and ‘Feb’ images from the ‘Lancaster’ Geodatabse to ArcMap. 

 

2.      Open the ‘image analysis’ window using ‘window’ option in ArcMap.

 

3.      We can set the ‘Image analysis options’ using  button on top left corner of the Image analysis window. Do it for both July and February images.

 

4.      Click the button. Under NDVI, make sure Red Band is 3 and Infrared Band is 4. Do it for both July and February images. This band combination holds true for Landsat Image. You might have other hypersprectral or satellite image where layer numbers will change. Also if you have stacked only two or three bands in your Composite image, Layer number could change accordingly.

 

 

5.      Select ‘Feb’ layer in Image analysis window and click on NDVI button  under processing section.

 

6.      You will see ‘NDVI_feb layer is created in ArcMap. Green color shows presence of vegetation and other colors show absence of green vegetation.

7.      Repeat the same process to get the July NDVI and save the data as ‘NDVI_July.

 

Questions:

1.      Turn in the image of Capitol Beach Lake with first order polynomial transformation option (Part I).

2.       Turn in output of NDVI for July and February for Lancaster County (Part II).  What is the prominent difference between July and Feb NDVI?

 

References

 

http://www.sc.chula.ac.th/courseware/2309507/Lecture/remote18.htm

http://www.youtube.com/watch?v=xVVdZOQiBuQ&feature=related

ArcGIS 10 help

ESRI training Visualizing and Analyzing Imagery with ArcGIS 10

Esri Training | Georeferencing Rasters in ArcGIS