Typical method of classification of remote sensing imagery has been pixel based. Normally, multispectral data are used to perform the classification and, indeed the spectral pattern present within the data for each pixel is used as the numerical basis for categorization. That is different feature types with different combination. Pixel based approach is based on conventional statistical techniques, such as supervised and unsupervised classification. In comparison to pixels, image objects carry much more useful information. Thus, they can be characterised by far more properties than pure spectral or spectral-derivative information, such as their form, texture, neighbourhood or context.”
The most evident difference between pixel-based and object-based image analyses is that first, in object oriented image analysis; the basic processing units are image objects or segments, not single pixels. Second, the classification in object oriented image analysis are soft classifies that is based on fuzzy logic. (Source 1, Source 2)
No more lectures, let’s start.
Today I am using landsat MSS image of river Jamuna of 1977. I will try to find out all the old sandbars along the river with Erdas 2011 Desktop. My pc has Dual Core 2.6gh 2mb processor and 4gb of RAM on Intel 31pr mainboard. First find the Imagine Objective at Raster tab of your version of Erdas. Open it.
1. You’ll meet a new window. Name your Project name and a New feature model.
2. In the Variable Properties window Add a new new variable which you want to work with. Then input the file path you desire to work with. Make sure the window matches the picture.
3. You are now at the Objective Workstation window.
Study the iconbar if your window does not look like mine.
4. Lets start the work. In the Project pane you’ll meet this tree. This is the area where we will add our logic and prerequisites. Don’t worry you won’t have to write codes!
5. Select Raster Pixel Processor (RPP). In the Properties tab select SFP as Available Pixel Cues. Hit the plus (+) sign next to it (see the above image).
6. The SFP will be added under RPP. Hit the ? mark next to + to know more about SFP. The help menu says,
Single Feature Probability (SFP) is a pixel cue that computes the probability metric (a number between 0 and 1) to each pixel of the input image based on its pixel value and the training samples. Higher probability values are assigned to those pixels whose values are similar to the ones of pixels in the non-background training samples. Lower probability values are assigned to pixels whose values are either similar to the ones of pixels in the background training samples or significantly different from the values of pixels in the non-background training samples.
7. Select SFP. At Training tab you’ll find a AOI toolbox pops up.
The magic starts here! pick some training pixel with the polygon tool and Add to the SFP Training area. Take 4 or 5 samples. Hit Accept.
8. Select SFP:<your file name>. In Distribution tab you can set the max and min pixel size.
9. Select Raster Object Creators (ROC), select Threshold and Clumps. No change here.