Multispectral Image Classification Based on Morphological Watershed Algorithm of a Remote Sensing Image Multidimensional Histogram

Abstract

Classical Multichannel image classification algorithms represent excessive segmentation and color intervention reflected from light on picture objects. Remote sensing image processing highly depends on the large data volumes often associated with multichannel imagery, including satellite imaging and aerial photos. This sort of data provides huge amount of information which is clearly arises the demand for automatic algorithms that are able to process these multi channel data sets fast enough for practical use and more accurately. This paper examines an improved image segmentation algorithm for the implementation of a multichannel classification procedure that brings together the information provided by all data channels. The proposed technique constructs multichannel morphological histograms, and deploys Radial Bases Function neural networks to define the morphological watershed markers that identify target datasets. The technique has been implemented on Landsat 9 multiband remotely sensed data.