I. Description of the problem
Used data includes Landcover, DEM, PM2.5 concentration, MODIS AOD;
you need to be match the scale and registration, in order to follow-up training modeling.
Second, the specific steps
-
The first step: Data Preview.
Recommended input data in the output data visualization ENVI, ArcGIS or in Matlab, and descriptive statistics, find and remove outliers, make accurate facilitate subsequent processing. -
Step two: image cropping.
Cut into the size of the study area, may be implemented in the ENVI. ENVI Classic 5.3 as follows: Basic Tools-Resize Data ---- In the window that pops up, select Input files, click Spatial Subset ---- pop-up window, you can cut a variety of ways to choose from, common is to click Map, enter the left and right corners of latitude and longitude, latitude and longitude range of cutting - click OK- set the output path and file name of each window - complete image cropping.
Sometimes problems image stitching, the same can be achieved by ENVI or ArcGIS, not repeat them here.
ENVI (Not ENVI Classic) : Start / Mosaicking / Seamless Mosaic in the Toolbox.
The ArcGIS : Providing Mosaic / Mosaic To New Raster functions in ArcToolbox ArcMap (ArcToolbox-Data Management Tools-Raster Mosaic / Mosaic To New Raster) -
A third step of resampling (uniform scale)
input and output video inverse problem should living same resolution, hence the need for resampling. ***
The main use of resampling tools include ENVI and MATLAB.
ENVIPrevious steps and similar, as follows: Basic Tools-Resize Data ---- In the window that pops up, select Input files, click OK ---- In the window that pops up, click Set Output Dims by Pixel Size ... - input resolution of the output image to be obtained, such as image resolution want 0.1 ° * 0.1 °, the input resolution of each window click 0.1,0.1- OK- output path and file name provided in the upper and lower block - complete resampling.
MatlabYou can imresize function. B = imresize (A, [mrows, ncols], method) A picture interpolation operation method specified by the parameter method, to change the size of the image to the aspect of [mrows, ncols]
Common resampling methodComprising: 'nearest' nearest neighbor interpolation, 'bilinear' bilinear interpolation, 'bicubic' bicubic interpolation, 'pixel average' average pixel (for downsampling / upscaling only). Methods need to consider the characteristics of the data. The land use types of data, using only nearest neighbor resampling.
Other, Outliers in the image will cause a larger image resampling, need to be addressed, with the exception modify or rewrite code resampling, no good approach, we have a good idea, also welcomed the sharing and exchange!
- A fourth step, the pixel alignment
after cutting and ENVI with resampling, even if the input and output images have the same spatial resolution and geographic scope, there may be problems with the pixel does not match, as follows:
This is because the cutting time ENVI and will not damage the pixel, i.e. in the figure, when the raster image to cropped blue region, an orange region will be, instead of the blue area.
This will cause problems pixels are still not aligned.
To solve this problem, you can use Clip tool ArcGIS when cropped cut. Specific operation is as follows:
first find the Clip tool in the toolbox, open:
in the pop-Clip window, enter the appropriate settings, click OK to:
the final result of cutting, and the ratio of the reference image, the pixels are completely aligned because were re-sampling, destroyed the original pixels:
So, when cut, it is recommended to use the ArcGIS Clip tool.
About pixel alignment, can further refer to this article: https://blog.csdn.net/kikitamoon/article/details/38254583
the above.
