pantierra/efast-phenocam-validation
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

changed to 3x3 window for index calc.

Browse source
This commit is contained in:
Felix Delattre 2026-02-08 20:16:11 +01:00
parent 8c03ae195d
commit 14a86f039a
2 changed files with 32 additions and 14 deletions
Download patch file Download diff file Expand all files Collapse all files

42
ndvi.py
View file

@ -87,14 +87,23 @@ def _get_ndvi_from_original(input_file, site_position):
if row < 0 or row >= src.height or col < 0 or col >= src.width: if row < 0 or row >= src.height or col < 0 or col >= src.width:
return None return None
r_val = float(red[row, col]) # Extract 3x3 window with boundary handling
n_val = float(nir[row, col]) r0, r1 = max(0, row - 1), min(src.height, row + 2)
c0, c1 = max(0, col - 1), min(src.width, col + 2)
red_window = red[r0:r1, c0:c1]
nir_window = nir[r0:r1, c0:c1]
if r_val <= 0 or n_val <= 0 or np.isnan(r_val) or np.isnan(n_val): # Calculate NDVI for each pixel in window
mask = (red_window > 0) & (nir_window > 0) & ~np.isnan(red_window) & ~np.isnan(nir_window)
if not np.any(mask):
return None return None
ndvi = (n_val - r_val) / (n_val + r_val) ndvi_window = np.zeros_like(red_window, dtype=np.float32)
return ndvi if not np.isnan(ndvi) else None ndvi_window[mask] = (nir_window[mask] - red_window[mask]) / (nir_window[mask] + red_window[mask])
# Return mean of valid NDVI values
valid_ndvi = ndvi_window[mask]
return float(np.mean(valid_ndvi)) if len(valid_ndvi) > 0 else None
except Exception as e: except Exception as e:
return None return None
@ -314,16 +323,25 @@ def _get_gcc_from_original(input_file, site_position):
if row < 0 or row >= src.height or col < 0 or col >= src.width: if row < 0 or row >= src.height or col < 0 or col >= src.width:
return None return None
b_val = float(blue[row, col]) # Extract 3x3 window with boundary handling
g_val = float(green[row, col]) r0, r1 = max(0, row - 1), min(src.height, row + 2)
r_val = float(red[row, col]) c0, c1 = max(0, col - 1), min(src.width, col + 2)
blue_window = blue[r0:r1, c0:c1]
green_window = green[r0:r1, c0:c1]
red_window = red[r0:r1, c0:c1]
total = r_val + g_val + b_val # Calculate GCC for each pixel in window
if total <= 0 or np.isnan(total): total = red_window + green_window + blue_window
mask = (total > 0) & ~np.isnan(total)
if not np.any(mask):
return None return None
gcc = g_val / total gcc_window = np.zeros_like(green_window, dtype=np.float32)
return gcc if not np.isnan(gcc) else None gcc_window[mask] = green_window[mask] / total[mask]
# Return mean of valid GCC values
valid_gcc = gcc_window[mask]
return float(np.mean(valid_gcc)) if len(valid_gcc) > 0 else None
except Exception as e: except Exception as e:
return None return None

4
run.py
View file

@ -24,7 +24,7 @@ def run_pipeline(season, site_position, site_name):
# print(f"Generating NDVI for raw data: {site_name}, {season}") # print(f"Generating NDVI for raw data: {site_name}, {season}")
# generate_ndvi_raw(season, site_position, site_name) # generate_ndvi_raw(season, site_position, site_name)
#create_ndvi_timeseries_raw(season, site_position, site_name) create_ndvi_timeseries_raw(season, site_position, site_name)
# print(f"Detecting clouds for {site_name}, {season}") # print(f"Detecting clouds for {site_name}, {season}")
#detect_clouds(season, site_name) #detect_clouds(season, site_name)
@ -40,7 +40,7 @@ def run_pipeline(season, site_position, site_name):
#process_cropped(season, site_position, site_name) #process_cropped(season, site_position, site_name)
#print(f"Generating NDVI for final outputs: {site_name}, {season}") #print(f"Generating NDVI for final outputs: {site_name}, {season}")
#generate_ndvi_post_process(season, site_position, site_name) #generate_ndvi_post_process(season, site_position, site_name)
#create_ndvi_timeseries_post_process(season, site_position, site_name) create_ndvi_timeseries_post_process(season, site_position, site_name)
#print(f"Generating GCC for final outputs: {site_name}, {season}") #print(f"Generating GCC for final outputs: {site_name}, {season}")
generate_gcc_post_process(season, site_position, site_name) generate_gcc_post_process(season, site_position, site_name)
create_gcc_timeseries_post_process(season, site_position, site_name) create_gcc_timeseries_post_process(season, site_position, site_name)