From bd46fb58093098cb7f39b03d69c14d5bbb89f971 Mon Sep 17 00:00:00 2001
From: Alexandre Bevington <alexandre.bevington@gov.bc.ca>
Date: Fri, 14 Aug 2020 16:50:34 -0700
Subject: [PATCH] Create raster_to_clean_polygon.R

First attempt at polygonizing and smoothing the model outputs
---
 R/raster_to_clean_polygon.R | 93 +++++++++++++++++++++++++++++++++++++
 1 file changed, 93 insertions(+)
 create mode 100644 R/raster_to_clean_polygon.R

diff --git a/R/raster_to_clean_polygon.R b/R/raster_to_clean_polygon.R
new file mode 100644
index 0000000..a3bd517
--- /dev/null
+++ b/R/raster_to_clean_polygon.R
@@ -0,0 +1,93 @@
+# SCRIPT TO POLYGONIZE, CLEAN, and SMOOTH RASTER OUTPUT (in tiles)
+# To do: add functionnality to stitch tiles using "summarize"
+# To do: compare smoothing results
+
+#  *****REQUIRES DEV VERSION OF STARS**** devtools::install_github("r-spatial/stars")
+
+library(smoothr)
+library(dplyr)
+library(stars)
+library(sf)
+
+#### RASTER TO POLYGON FUNCTION ####
+
+# READ CLASSIFIED RASTER
+  r <- stars::read_stars('data/out/3CategoryPrediction/20190926-103025_map_recl.tif', proxy = F)
+  d <- st_dimensions(r)
+  int <- 1000
+  x_seq <- seq(d$x$from,d$x$to+int,int)
+  
+# CLEANING FUNCTION (PER TILE)
+  ras_to_clean_pol <- function(seq_along = 1,
+                               out_folder = "data/clean/",
+                               out_format = "kml",
+                               wetland_class_number = c(2),
+                               min_area_drop_m2 = 625 * 4,
+                               max_area_fill_m2 = 625 * 4,
+                               raw_out = F,
+                               clean_out = F,
+                               smooth_out = T,
+                               smooth_method = "chaikin",
+                               return = T
+                               ){
+    
+    write.csv("", paste0(out_folder,seq_along,"_start.txt"))
+    # CLIP MY TILE
+    my_tile <- r[1,x_seq[seq_along]:x_seq[seq_along+1],] %>% st_as_stars()
+    # print("ok subset")
+    
+    # POLYGONIZE (AND DISSOLVE/MERGE BY CLASS)
+    p <- sf::st_as_sf(my_tile,
+                      as_points = F,
+                      merge = T)
+    # print("ok poly")
+    
+    # RENAME FIELDS
+      names(p) <- c("class","geometry")
+  
+    # SELECT WETLANDS OF INTEREST
+    p <- p %>%
+      filter(class %in% wetland_class_number)
+  
+    # CALCULATE AREA (m2) (based on trim 25x25 data, so 625 m2 is one pixel)
+    p <- p %>%
+      mutate(area = st_area(.))
+    # print("ok area")
+    
+    if(raw_out == T){
+      write_sf(p, paste0(out_folder,seq_along,"_raw.",out_format))
+      # print("ok raw")
+      }
+  
+    # FILTER AREA (i.e. DROP CRUMBS)
+    p <- p %>%
+      filter(area > units::as_units(min_area_drop_m2, "m2"))
+      # print("ok crumbs")
+      
+    # FILL HOLES
+    p <- p %>%
+      fill_holes(threshold = units::as_units(max_area_fill_m2, "m2"))
+      # print("ok holes")
+      
+    # WRITE CLEAN
+    if(clean_out == T){
+      write_sf(p, paste0(out_folder,seq_along,"_clean.",out_format))
+      # print("ok clean")
+      }
+    
+    # SMOOTH POLYGONS
+    if(smooth_out == T){
+      p_sm <- smooth(p, method = smooth_method)
+      write_sf(p_sm, paste0(out_folder,seq_along,"_smooth.",out_format))
+      # print("ok smooth")
+      }
+    
+    if(return == "clean"){
+      return(p)
+    }
+    if(return == "smooth"){
+      return(p_sm)
+    }
+  }
+
+pol <- do.call(rbind, lapply(1:length(x_seq), ras_to_clean_pol))