Use of machine learning methods forsuperpixel-based spatial classification and segmentation

Engineer's thesis code supplement

In this repistory one can find the implementation of ideas presented in my engineering thesis together with the data needed to recreate the results.

All scripts can be found in the scripts folder:

• install_packages.R - run the code in this script to install all the required packages
• source.R - script with functions used in other scripts, mostly technical
• lrs_hrs.R - script with the implementation of my method
• smst.R - script with the implementation of the SMST method

In both lrs_hrs.R and smst.R the image to be processed needs to be chosen. All values needed for both approaches are stored in the choose_image data frame.

                      file smooth_threshold lrs_segments hrs_segments smst_segments
2017-05-19 orto_170519.tif               40           24           54            53
2018-04-07 orto_180407.tif               60           27           78            44
2019-08-31 orto_190831.tif               50           10           87            49
2020-03-28 orto_200328.tif               45           12           51            42
2020-04-06 orto_200406.tif               50            7           76            67
2021-10-01 orto_211001.tif               55           13           64            64

Choosing the image is done by supplying the image date or image number for row extraction from the choose_image data frame:

current_image = choose_image["image-date",]

#or 

current_image = choose_image[image_number,]

As an example:

current_image = choose_image["2017-05-19",]

#or 

current_image = choose_image[1,]