Generate-and-align-orthomosaics

---

Summary

The generate and align orthomosaics article describes the process to generate aligned orthomosaics from repeated UAV survey.

Fig 1 : Generate and align orthomosaics

---

Requirements

This step-by-step guide requires :
- canObsR well installed, the conda environment well configurated and the Metashape license activated. Check the installation guide
- Repeated UAV surveys database well structured. Check the prepare database article to organize your data or download and unzip ‘canObsR_data.zip’.Test data contains, 3 sets of drones images from the same area obtained from 3 different dates and a CHM canopy height model from the area which is the reference.

---

Step-by-step exemple

Here is a short example to guide you step by step through the orthomosaic generating process and their alignment with the functions generate_Mosa() and align_Mosa(). These functions required, arosics(AROSICS), the Metashape python API and a valid Metashape license.

library(canObsR)
library(reticulate)

use_condaenv('canObsR-env')

my_path <- 'XXXX/canObsR_data'

generate_Mosa(path_in = file.path(my_path, '1_my_drone_data'),
          out_dir_ortho = file.path(my_path, '2_orthomosaics'),
          data_type = "RGB",
          resol_ref = 0.05,
          site_name = "Bouamir",
          crs = "EPSG:32633",
) # Write orthomosaics in 2_orthomosaics folder

align_Mosa(path_in = file.path(my_path, '2_orthomosaics'),
        ref_filepath = file.path(my_path, 'Bouamir_LiDAR_ref.tif'),
        out_dir_path = file.path(my_path, '3_orthomosaics_aligned'),
        corr_type = "global",
        #grid_res = 200,
        window_size = 500,
        window_pos = list(258131, 352973),
        apply_matrix = TRUE,
        save_data = FALSE
) # Write aligned orthomosaics in 3_orthomosaics_aligned folder

Non dynamic process of generating and aligning orthomosaics with test data

Additional utils

When aligning a long time serie of mosaics, you may want to dynamically register each mosaic to the previous one instead of keeping one reference for the whole serie, in order to limit the time difference between a mosaic and its reference.

Here follows a snippet of code that allows you to do just that : the first image is aligned using the specified reference file, and then each subsequent image N uses the aligned version of the image N-1 as a reference.

dir_mosa = file.path(my_path, '2_orthomosaics')
out_dir_path = file.path(my_path, '3_orthomosaics_aligned')


PATHs.df = data.frame(
   path.in = dir(dir_mosa, full.names = T),
   output_mosa_name = basename(dir(dir_mosa, full.names = T)),
   path_ref = NA
)

# Add reference used for each mosaic
for(i in 1:nrow(PATHs.df))                      
{
  if (i==1){
     PATHs.df$path_ref[i] = file.path(my_path, 'Bouamir_LiDAR_ref.tif')
  }else{
     PATHs.df$path_ref[i] = paste0(out_dir_path, PATHs.df$output_mosa_name[i-1]) 
  }
}

# Do the alignement with the dynamic alignment process
for(i in 1:nrow(PATHs.df)) 
{
  path_in = PATHs.df$path_in[i]
  ref_filepath = PATHs.df$path_ref[i]
  
  align_Mosa(path_in = path_in,
        ref_filepath = ref_filepath,
        out_dir_path = out_dir_path,
        corr_type = "local",
        grid_res = 200,
        window_size = 500,
        apply_matrix = FALSE,
        save_data = TRUE
        )
 
  print("<!> DONE <!>")
  print("######################################")
}

Dynamic process of generating and aligning orthomosaics with test data

Note : In this case, both the input and reference images are potentially heavy mosaics. Depending on their size, it may be useful to downgrade their resolution to get a faster result without much impact on quality.

It is also possible to save the transform applied to an image, and to later apply it to another one, using the apply_saved_matrix() function.

my_path <- 'XXXX/canObsR_data'

apply_saved_matrix(im_path = file.path(my_path, '2_orthomosaics', 'additional_ortho.tif'),
                   out_dir_path = file.path(my_path, '3_orthomosaics_aligned'),
                   metadata_path = file.path(my_path, '3_orthomosaics_aligned', 'Bouamir_20220511_metadata.pkl') 
                   )

Note : Use this only if both images have the same extent.