andreybicalho/extendedmorphologicalprofiles

Remote sensed hyperspectral image classification with Spectral-Spatial information provided by the Extended Morphological Profiles

Remote Sensed Hyperspectral Image Classification With The Extended Morphological Profiles and Support Vector Machines

This is an example of how to use the Extended Morphological Profiles and Support Vector Machines to classify remote sensed hyperspectral images using Python.

Indian Pines Dataset

This scene was gathered by AVIRIS sensor recorded over Northwestern Indiana, USA, and consists of 145x145 pixels and 224 spectral reflectance bands in the wavelength range 0.4–2.5 10^(-6) meters. The Indian Pines scene contains two-thirds agriculture, and one-third forest or other natural perennial vegetation. There are two major dual lane highways, a rail line, as well as some low density housing, other built structures, and smaller roads. The ground truth available is designated into sixteen classes (seventeen if you consider the background) and is not all mutually exclusive. It is also a very common practice reducing the number of bands to 200 by removing bands covering the region of water absorption: [104-108], [150-163], 220. Indian Pines data are available through Pursue's univeristy MultiSpec site.

Extended Morphological Profiles (EMP)

The Extended Morphological Profiles (EMP) is a simple and effective technique to encode both spectral and spatial information in the classification process. This method connects similar structures through morphological operations and keeps the essential spectral information by using some feature-extraction method such as Principal Component Analysis (PCA).

Classification: Support Vector Machines (SVM)

In this example the Support Vector Machine (SVM) machine learning algorithm, with the Radial Basis Function (RBF) Kernel, was used for the classification.

extendedmorphologicalprofiles's Issues

The paper you followed

hello, your code is perfect, but i would like to know which article you are referring to, thanks!

Thank you

Thank you it is a good code.
Do you know any similar open source code(EMPs)? I'm curious how/where the authors implemented their method in papers.

problems with low accuracy

I'm getting a really low accuracy using 10% training data on Indian pines ( getting 31% OA ) although the paper says they're getting 93% OA.
even when using the default values you are using, i'm getting 40% OA.