Unlike pixel-based classification methods, which classify individual pixels directly, object-based classification aggregates image pixels into spectral homogeneous image objects using an image segmentation algorithm before classifying the individual objects.
The distinctions between object-based and pixel-based classification methods can be seen from two perspectives: classification units and classification features.
A land cover class is assigned to each pixel in traditional pixel-based image classification. All pixels are the same size and shape, and they have no idea who their neighbors are.
OBIA, on the other hand, segments an image by grouping small pixels together into vector objects. Instead of digitizing the image on a per-pixel basis, segmentation does it for you.
there are two processes those are segmentation and classification performed in Object-based Image Analysis (OBIA).
What is segmentation ?
it is key to classification in object based
Refers to grouping pixels to form objects.
It is a process of grouping the pixel with the same brightness level or gray scale to make an image clear for object based classification
In segmented object , you use their spectral, geometrical, and spatial properties to classify them into land cover.
What is classification ?
It is the process of sorting the objects based on their shape, size, spatial, and spectral properties.
classification process involves the following are brief descriptions of two common classification methods. There is no “best” method or combination of methods, just as there is no “best” segmentation process. The most appropriate method is determined by the user’s objectives, image characteristics, prior knowledge, experience, and preferences.
Nearest neighbor (NN)
1. The user selects sample image objects for each class.
2. Samples are typically selected based on prior knowledge of the plant community and should represent the range of characteristics within a single class.
3. The software searches for objects that are similar to the samples and then assigns those objects to the appropriate class.
4. Iterative steps improve classification.
5. Appropriate for describing variation in images with high resolution.
Membership function
The
- The user selects features with different value thresholds for different classes.
- The software categorizes image objects based on the feature threshold set by the user (see example below)
- The results are more objective than NN, and they are simple to edit.
- Useful if the classes can be easily distinguished by a single or a few characteristics.
- When there is little prior knowledge about the particular vegetation community in the image, this method is appropriate.
What is the Advantage of using OBIA?
A. Multiple scales
Image objects contain spatial relationship information that allows for more than one level of analysis. This is critical because image analysis at the landscape scale necessitates a number of related levels of segmentation, or scale levels.
B. Spatial relationships
Objects can be classified based on their spatial relationships with other objects.
C. Information filter
OBIA is capable of filtering out meaningless data and assimilating it with other data into a single object. This is similar to how the human eye filters information, which is then translated by the brain into a meaningful image.
D. Output
OBIA typically produces a classified image, which is then used as part of a map, for example, to illustrate different vegetation types in a given area. The segmentation itself can be an output and is frequently imported as a raster into a GIS.
