I have Pavia dataset (.mat file ; hyperspectral image) which i downloaded from "http://www.ehu.eus/ccwintco/index.php?title=Hyperspectral_Remote_Sensing_Scenes". There are classes as water, tree, titles etc. Here i want to know how to exact these samples from the dataset & from which parameter they differ to each other.
You can follow following steps in envi software:
per-processing of your data.
identify land use classes using visual interpretation of images of taking help of google earth
make ROI for each class and name accordingly.
go for the classification algorithm give in the list of ENVI classifiers
select your reflectance (preprocessed) hyper-spectral image as input.
define your ROIs as trining data.
classify.
any further query is most welcome.
best wishes
Also, Matlab editor is a proper environment to extract mat files for a data set.
I think the clustering algorithm works very well to group the data set.There are so many clustering like k-means,FCM and many more algorithm.These clusters group the dataset.To find the parameter u extract some features from these classes and take the mean of it and plot.The features which shows distinct variation shall be considered.
I used two clustering algorithms (including K-means and fuzzy c-means) in my article and conference papers about automatic building extraction through high resolution satellite multi-spectral imagery. They help you to extend these approaches in hyper-spectral imagery.
dear shrish
if you want to good accuracy classified image, you need to collect ground truth data,then you have preprocessing the data of satellite product, you find some standard operational procedure,you have to see in envi hyperspectral tutorials.
best wishes
- Badges
- Science topic
- Similar topics
- Geoscience
- Atmospheric Sciences
- Atmosphere