0

I am (very) new to GIS in general. I would like to convert a satellite image of a province into raster, assigning each pixel its own data (i.e., a number and a name) based on its html color code. For example, shades of red and/or blue would be assigned the data "id=1; name=Residential" and shades of light green would be "id=2; name=grassland" and green to dark green would be "id=3; name=forest".

I'm thinking of taking screen shots (zoomed in as much as possible) and then combining them to capture the whole province. Next, I'll georeference the image.

How could I go about this next?

PolyGeo
  • 65,136
  • 29
  • 109
  • 338
HobbitLass
  • 1
  • 1
  • 1
    what software are you using? – MDHald Sep 02 '15 at 14:13
  • 1
    If you have a satellite image, it is already a raster. It sounds like you are wanting to to perform image classification. – Michael B Sep 02 '15 at 14:16
  • QGIS for now but I'm considering trying ArcGIS 10.2 Student Edition (trial) if it would be more convenient to the cause. :)

    Someone suggested creating polygons first and then converting into raster but it's taking forever; 1110 polygons as of today but I'm far from done. Since most of the province is made up of dark green trees, light green grassland, brown soil, grey roads, and white/blue/red rooftops, I was hoping converting the satellite images would be faster.

    When I said satellite image, I meant images via Google Earth. I'm using the openlayers plugin in QGIS as base for the polygons.

     – HobbitLass Sep 02 '15 at 14:18
  • since you want to avoid the vector to raster switch I would say to train the images using PCI Geomatica...However! there is a cost to buy the software (even as a student). If you do find a copy here is a Tutorial – MDHald Sep 02 '15 at 14:25
  • Screenshots of google earth images??? That sounds very unscientific! – Iris Sep 02 '15 at 14:25

1 Answers1

1

It sounds like you will want to be completing Supervised Classification (I can't find any reliable articles to explain the theory behind Supervised Classification). Take a look at the Supervised Classification to Create Vegetation Layer and Supervised and Unsupervised Classification in ArcGISto see how to use the Image Analysis tool in ArcGIS. If you have the Spatial Analyst Extension you can watch the video Supervised Image Classification and they will guide you through it.

MDHald
  • 1,871
  • 13
  • 20
  • oops one more thing - as Supervised Classification will create a Vector format you will need to convert it to Raster using the GDAL Rasterize tools. Take a look at the article How to convert a vector layer to raster? – MDHald Sep 02 '15 at 14:22
  • Thank you for your response, MDHald. I'd love to give it a go right ahead but I need to first get a hold of a 3-band and 4-band imagery of this particular province in the Philippines. I'll let you know how it goes. – HobbitLass Sep 02 '15 at 14:28
  • I don't know what province you need but you can use the Mapbox Landsat Live Map to get a list of your imagery. Gather the file number (EG.LC80150332015229LGN00) then go to AWS Landsat Live using the link format http://landsat-pds.s3.amazonaws.com/L8/015/033/LC80150332015229LGN00/index.html

    Replace 015 with the Path and replace 033 with the Row. Then gather the title _B3 and _B4 from the link

     – MDHald Sep 02 '15 at 14:56
  • you can download different image data (some are free, some not) here: http://earthexplorer.usgs.gov/ – Iris Sep 02 '15 at 15:17
  • @MDHald: It's the Guimaras, Philippines. Thank you for the possible (and hopefully free) imagery source. I'll do my best to follow the instructions. – HobbitLass Sep 04 '15 at 12:06
  • @Iris: I appreciate having another imagery source, thank you. There are a lot of data set types, most of which I'm seeing for the first time. I'm going to need a bit of hand holding to figure out which exactly fits my purpose. For now, I'm going to try each data set filter to check its availability. – HobbitLass Sep 04 '15 at 12:06