I find HSI images on this link: http://www.ehu.eus/ccwintco/index.php?title=Hyperspectral_Remote_Sensing_Scenes. But all are having more than 10 no. of channels. Want to know how can I use this in matlab for image classification. Because I don't know what are the RGB channels among various channels. If it is possible convert them into tiff format? Thanks in advance
http://www.ehu.eus/ccwintco/index.php?title=Hyperspectral_Remote_Sensing_Scenes
Hi Regan,
Matlab does not directly support Tiff outputs with more than 3 bands, as you say. You can select the RGB channels from the hyperspectral image by checking the specific sensor's information and creating an image with 3 bands by choosing a band in the red, green and blue spectral ranges respectively. Anyway by doing so you are throwing away a lot of useful spectral information in the process. To handle hyperspectral images in Matlab I would suggest instead to install the Hyperspectral Toolbox, which includes read/write/visualization routines, available at: http://sourceforge.net/projects/matlabhyperspec/
Hope that helps!
If you want to make classification, it does not matter which is the number of channel. You can have an image of NxM as size with 10 channels. It will just imply that the number of feature dimension will be 10. So, you will need to construct a matrix with N by M row and 10 column (if 10 channels). Use 90 % of the row for training and test with the rest for instance with a 10-fold cross validation for instance.
The aforementioned toolbox should allow you to visualise the data.
- Badges
- Science topic