Why an Object-Oriented Approach?
Before we answer the question “Why object-oriented technology?” we will discuss the differences between pixel-based versus object-based image analyses using a practical application of multispectral remote sensing imagery. Let us take a look at the example of feature extraction of airport runways from imagery to demonstrate the point. Runways usually contain a variety of spectral signatures comprised of bright white paint used for runway markers and numbers, concrete runways, and black tread marks of planes landing. Traditional pixel-based supervised classification approaches require the creation of training sets to include the different spectral signatures of various features within the runway. Considering the advantages of an object-oriented approach to runway extraction, imagine if we somehow managed to create objects that defined the runway and features contained within the runway, then the problem would be simplified to classifying a few polygons rather than hundreds of pixels that are within the runway. Also, we can take advantage of the shape features, such as minimum size of the objects, to eliminate island pixels, that is, pixels that are entirely different as compared to the surrounding pixels. Also, we can use the same supervised classification technique used on pixels to classify the objects. Figure 3.1 demonstrates the results of feature extraction using the objectoriented approach.