With urbanization and urban sprawl, land-use and land-cover
change studies became an area of keen interest for researchers. Cause of rapid
urbanization is peoples’ responses to economic opportunities.
In today’s tutorial a variety of homogeneous and heterogeneous objects were found with in small areas. Urban features of variable sizes, shapes, inconsistent structures and layout/patterns were identified Spectral diversity and similarities of the urban features due to different types of construction material was found. Trees were problematic because they were found covering roads, streets, and building roofs. Shadow caused by buildings and trees were hiding actual ground features
 |
| Image by Alexas_Fotos from Pixabay |
“Image Classification
is the process of sorting pixels into
a finite number of individual classes, or categories, of data based on their
data file values”. If a pixel satisfies a certain set of criteria, then the
pixel is assigned to the class that corresponds to that criteria. Remote
Sensing generally classify data by Supervised & Unsupervised methodology
Supervised classification is more closely controlled by analyst
than unsupervised classification. In this process, analyst select pixels that represent
patterns that can be identified with help from other sources. Knowledge of the
data, the classes desired, and the algorithm to be used is required before you
begin selecting training samples. By identifying patterns in the imagery you
can "train" the computer system to identify pixels with similar
characteristics. By setting priorities to these classes, analyst can supervise
the classification of pixels as they are assigned to a class value. If the
classification is accurate, than each resulting class corresponds to a pattern
that you originally identified.
With advent of modern high resolution era of
satellite images, PIXEL based, Object based and many other machine learning
algorithms have been developed. This
assignment will use high resolution image and will apply supervised pixel based
methodology to asses a certain land cover.
Methodology
- Start ERDAS Imagine, From the Viewer Tool Bar select the Open Raster Layer icon and select the file
- Display as Fit to Frame.
- From the Main Icon Panel select Raster-> Supervise-> Signature Editor. A dialog box will appear and will eventually contain a Cell Array of created signatures.
- Go to drawing tab -> Draw a polygon ->
- The area selected will now be taken as a sample for the classifier. From the Signature Editor, Click add Button, signature will appear in editor window.
- Next take many samples all over the image and select all holding shift key as shown now click Add again.
- Next select all signatures by click from 1-till last and click merge a new class appears rename this to e.g. residential A and delete all others by right clicking and selecting Delete Selection.
- Once you are done From the Signature Editor Menu Bar select File/Save. Name your file filename.sig. Click OK.
- Next I went to raster tab-> Supervised Classification. Specify inputs and outputs
- For the Parametric Rule I selected Maximum Likelihood...
- When the process is done click OK in the Job Status box. It will display classified.img, and the original image side by side in viewers.
- Next I opened Arc Map for map making and adjusting classes where required
Conclusion
In today’s tutorial a variety of homogeneous and heterogeneous objects were found with in small areas. Urban features of variable sizes, shapes, inconsistent structures and layout/patterns were identified Spectral diversity and similarities of the urban features due to different types of construction material was found. Trees were problematic because they were found covering roads, streets, and building roofs. Shadow caused by buildings and trees were hiding actual ground features
Technique used in the lab cannot separate most of the
spectrally similar classes, and produce non-homogeneous classes due to internal class
spectral variability, because of high resolution data.
Classification over such areas in a high resolution images
produces speckled results, salt and pepper effects.
References
- E. Mendoza, R. Dirzo Deforestation in Lacandonia (southeast Mexico): evidence for the declaration of the northernmost tropical hot-spot
- Biodiversity and Conservation, 8 (1999), pp. 1621-1641
- B. Mertens, W. Sunderlin, O. Ndoye, E.F.Lambin Impact of macro-economic change on deforestation in South Cameroon: integration of household survey and remotely-sensed data
- World Development, 28 (2000), pp. 983-999
- E.F. Moran Deforestation and land use in the Brazilian Amazon
- Human Ecology, 21 (1993), pp. 1-21
- Working the Sahel: Environment and society in Northern Nigeria, Routledge, London (1999)
- M. Mortimore, M. Tiffen Population growth and a sustainable environment
- G. Oba, N.C. Stenseth, W.J. LusigiNew perspectives on substainable grazing management in arid zones of sub-Saharan Africa
- E. Ostrom, J. Burger, C.B. Field, R.B.Noorgaard, D. Policansky Revisiting the commons: Local lessons, global challenges
Comments
Post a Comment