An Object Based Image Analysis (OBIA) approach to image classification segments an image into homogenous spectral regions. Image segmentation is the process of grouping multiple adjacent pixels into polygon ‘Objects’ based on similar spectral characteristics. This differs from standard image analysis methods that classify each pixel in isolation; the OBIA approach generally produces a cleaner map product, more meaningful to human users.
Using the new ‘chain tools’ function in SAGA, that allows multiple modules to be linked into one process, the developers have created a new Object Based Image Analysis module that works beautifully.
Below is an example of using this function for classify a mine area in Sulawesi, Indonesia.
Another example – mapping fires in Northern Australia. – Updated April 2015
This example uses SAGA-GIS 2.2.5 to conduct a supervised OBIA segmentation based classification.
The example below shows how to map one fire event by subtracting an image after a fire from one before. The example is of a fire that escape from a mine tenure into Kakadu national park.
NOTE: I have noticed on one install the Tool Chain modules missing including the OBIA classification. This can be easily fixed by going to saga_2.2.0_x64\modules\toolchains and dragging and dropping the tool chain modules into SAGA.