Search the Community

I am using Erdas Imagine 2015 software to perform an unsupervised classification of a Landsat 8 img file. When using the Isodata method is there any parameters I can adapt such as standard deviation, convergence threshold etc that are a standard to produce the most accurate results for a land cover assessment rather than the default settings. As recoding and masking individual classes is a lengthy and complicated process. Thanks

Hi GIS Area, I just made this account wondering if is anybody here could help me making a simple classification on high resolution satelite imagery (GeoEye). The problem is that the parameters in the software are not so clear for me, both, unsupervised and supervised, and I haven't found any conclusive solution on the web. The methods I've tried are the K-mean, which gives me a highly fragmented image, and also ISOData, the parameters were all executed in the default, the only thing I did was to change classes from 5 to 3, or in the ISOData case, 3 or 4. The supervised class. asks for a ROI and EVF file, which I searched on the internet and didn't understand what was going on with these files. What I did was to make some polygons in the ArcMAP 10.0 and did a supervised classification just there, again, with a very fragmented result. I don't think it is a very hard situation, my objective here is just to divide the landscape into 2 or 3 main "clusters" which is pasture, forestry, and in some cases, water bodies (as is the case of the image posted in the topic) The softwares I have available are ENVI 4.7 and 5.0. If someone could give me at least a hint or something like that, I would be really appreciate. The satelite image is available for everyone in the google drive in the following link: https://drive.google.com/file/d/0B1zf2D-uJw4wYWxQLWh0MFhZa3c/edit?usp=sharing Regards! Raôni Zanovello