You cannot select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
442 lines
17 KiB
Python
442 lines
17 KiB
Python
# python 3.5
|
|
#requires LAStools to be installed (with the appropriate license). Note that LAStools requires no spaces in file names
|
|
|
|
#should have previously run 2017088_las_manipulation to have a las that has the buildings and veg removed
|
|
#note that the neilson volumes script must be in the same folder
|
|
|
|
# this script will:
|
|
#crop to a given polygon (crop away the swash zone)
|
|
# extract values along a predefined profile,
|
|
# do the volume analysis
|
|
#export pngs of the surveys
|
|
########################### IMPORTS ###########################################
|
|
import os
|
|
import subprocess
|
|
import pandas as pd
|
|
import numpy as np
|
|
import neilson_volumes
|
|
import matplotlib.pyplot as plt
|
|
from matplotlib.ticker import MultipleLocator
|
|
import datetime
|
|
import xlsxwriter
|
|
import math
|
|
from cycler import cycler
|
|
|
|
###############################################################################
|
|
########################## FIXED INPUTS #######################################
|
|
######### UNCOMMENT THIS SECTION IF YOU WANT TO DEFINE EACH INPUT INDIVIDUALLY ######################
|
|
##path to LasTools NOTE THERE CAN BE NO SPACES
|
|
path_2_lastools='C:/ProgramData/chocolatey'
|
|
# input_file=r"J:\Project\wrl2017088 Central Coast Council Aerial Survey and Coastal Analysis\04_Working\Python\Survey 2\Parameter Files\las outputs survey2.xlsx"
|
|
input_file='Parameter Files/las outputs survey2.xlsx'
|
|
#input_file=r"J:\Project\wrl2017088 Central Coast Council Aerial Survey and Coastal Analysis\04_Working\Python\Survey 2\Parameter Files\las outputs survey2_V2.xlsx"
|
|
##############################
|
|
############################### SUB ROUTINES ##################################
|
|
def check_output(command,console):
|
|
|
|
if console == True:
|
|
process = subprocess.Popen(command)
|
|
else:
|
|
process = subprocess.Popen(command, shell=True, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, universal_newlines=True)
|
|
output,error = process.communicate()
|
|
returncode = process.poll()
|
|
|
|
return returncode,output
|
|
|
|
def crop_las(las, shp, output, lastools_loc):
|
|
# output is the full path and filename (inc extension) to put in
|
|
path_2_lasclip=lastools_loc+"\\bin\\lasclip"
|
|
|
|
command="%s -i %s -poly %s -o %s" % (path_2_lasclip, las, shp, output)
|
|
|
|
returncode,output = check_output(command, False)
|
|
|
|
if returncode!= 0:
|
|
print("Error. lasclip failed on %s" % shp.split('//')[-1].split('.')[0])
|
|
else:
|
|
return None
|
|
|
|
def las_boundary(las, crop_poly_name, path_2_crop_polygon, lastools_loc, zone):
|
|
path_2_lasboundary=lastools_loc+"\\bin\\lasboundary"
|
|
fname=crop_poly_name
|
|
prjfname="%s%s.prj" %(path_2_crop_polygon, fname)
|
|
path_2_crop_poly='%s%s.shp' % (path_2_crop_polygon, fname)
|
|
command="%s -i %s -o %s" % (path_2_lasboundary, las, path_2_crop_poly)
|
|
|
|
prj=open(prjfname, 'w')
|
|
if zone==56:
|
|
prj.write('PROJCS["GDA_1994_MGA_Zone_56",GEOGCS["GCS_GDA_1994",DATUM["D_GDA_1994",SPHEROID["GRS_1980",6378137.0,298.257222101]],PRIMEM["Greenwich",0.0],UNIT["Degree",0.0174532925199433]],PROJECTION["Transverse_Mercator"],PARAMETER["False_Easting",500000.0],PARAMETER["False_Northing",10000000.0],PARAMETER["Central_Meridian",153.0],PARAMETER["Scale_Factor",0.9996],PARAMETER["Latitude_Of_Origin",0.0],UNIT["Meter",1.0]]')
|
|
elif zone==55:
|
|
prj.write('PROJCS["GDA_1994_MGA_Zone_55",GEOGCS["GCS_GDA_1994",DATUM["D_GDA_1994",SPHEROID["GRS_1980",6378137.0,298.257222101]],PRIMEM["Greenwich",0.0],UNIT["Degree",0.0174532925199433]],PROJECTION["Transverse_Mercator"],PARAMETER["False_Easting",500000.0],PARAMETER["False_Northing",10000000.0],PARAMETER["Central_Meridian",147.0],PARAMETER["Scale_Factor",0.9996],PARAMETER["Latitude_Of_Origin",0.0],UNIT["Meter",1.0]]')
|
|
elif zone==57:
|
|
prj.write('PROJCS["GDA_1994_MGA_Zone_57",GEOGCS["GCS_GDA_1994",DATUM["D_GDA_1994",SPHEROID["GRS_1980",6378137.0,298.257222101]],PRIMEM["Greenwich",0.0],UNIT["Degree",0.0174532925199433]],PROJECTION["Transverse_Mercator"],PARAMETER["False_Easting",500000.0],PARAMETER["False_Northing",10000000.0],PARAMETER["Central_Meridian",159.0],PARAMETER["Scale_Factor",0.9996],PARAMETER["Latitude_Of_Origin",0.0],UNIT["Meter",1.0]]')
|
|
prj.close()
|
|
|
|
returncode,output = check_output(command, False)
|
|
|
|
return None
|
|
|
|
def make_raster(las, output, lastools_loc, keep_only_ground=False, step=0.2):
|
|
#not that keep ground only option only rasters points classified as "2" in the lidar (ie ground)
|
|
#this effectively creates a "bare earth dem"
|
|
#note that this should only be used after remove_veg and/or remove_buildings has been run
|
|
|
|
path_2_blastdem=lastools_loc+"\\bin\\blast2dem"
|
|
|
|
if keep_only_ground==False:
|
|
command="%s -i %s -o %s -step %s" % (path_2_blastdem, las, output, step)
|
|
else:
|
|
command="%s -i %s -o %s -step %s -keep_class 2" % (path_2_blastdem, las, output,step)
|
|
|
|
returncode,output2 = check_output(command, False)
|
|
|
|
if returncode!= 0:
|
|
print("Error. blast2dem failed on %s" % las.split('\\')[-1].split('.')[0])
|
|
else:
|
|
return None
|
|
|
|
|
|
def extract_pts(las_in, cp_in, survey_date, keep_only_ground=True):
|
|
"""Extract elevations from a las surface based on x and y coordinates.
|
|
|
|
Requires lastools in system path.
|
|
|
|
Args:
|
|
las_in: input point cloud (las)
|
|
cp_in: point coordinates with columns: id, x, y, z (csv)
|
|
survey_date: survey date string, e.g. '19700101'
|
|
keep_only_ground: only keep points classified as 'ground' (boolean)
|
|
|
|
Returns:
|
|
Dataframe containing input coordinates with extracted elevations
|
|
"""
|
|
|
|
cmd = ['lascontrol', '-i', las_in, '-cp', cp_in, '-parse', 'sxyz']
|
|
|
|
if keep_only_ground == True:
|
|
cmd += ['-keep_class', '2']
|
|
|
|
# Call lastools
|
|
process = subprocess.Popen(
|
|
cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
|
|
stdout, stderr = process.communicate()
|
|
errcode = process.returncode
|
|
|
|
# Handle errors, if detected
|
|
if errcode != 0:
|
|
print("Error. lascontrol failed on {}".format(
|
|
os.path.basename(las_in)))
|
|
print(stderr.decode())
|
|
|
|
# Load result into pandas dataframe
|
|
df = pd.read_csv(io.BytesIO(stdout))
|
|
|
|
# Tidy up dataframe
|
|
df = df.drop(columns=['diff'])
|
|
df['lidar_z'] = pd.to_numeric(df['lidar_z'], errors='coerce')
|
|
df['Beach'] = beach
|
|
df = df[[
|
|
'Beach', 'ProfileNum', 'Easting', 'Northing', 'Chainage', 'lidar_z'
|
|
]]
|
|
|
|
# Rename columns
|
|
new_names = {
|
|
'ProfileNum': 'Profile',
|
|
'lidar_z': 'Elevation_{}'.format(survey_date),
|
|
}
|
|
df = df.rename(columns=new_names)
|
|
|
|
return df
|
|
|
|
|
|
|
|
def process_tmp_csv(tmp_csv_file, survey_date, out_csv_loc, beach_name):
|
|
#lascontrol does not output in a particularly useful format.
|
|
#this processes its output to something useful
|
|
|
|
tmp_csv=pd.read_csv(tmp_csv_file)
|
|
|
|
profile=tmp_csv['ProfileNum']
|
|
z=tmp_csv['lidar_z']
|
|
x=tmp_csv['Easting']
|
|
y=tmp_csv['Northing']
|
|
chain=tmp_csv['Chainage']
|
|
|
|
data={}
|
|
|
|
i=0
|
|
while i<len(profile):
|
|
profile_now=profile[i]
|
|
data[profile_now]={}
|
|
easting=[]
|
|
northing=[]
|
|
chainage=[]
|
|
elevation=[]
|
|
while i<len(profile) and profile[i]==profile_now:
|
|
easting.append(x[i])
|
|
northing.append(y[i])
|
|
elevation.append(z[i])
|
|
chainage.append(chain[i])
|
|
i=i+1
|
|
data[profile_now]['Easting']=easting
|
|
data[profile_now]['Northing']=northing
|
|
data[profile_now]['Elevation']=elevation
|
|
data[profile_now]['Chainage']=chainage
|
|
|
|
for profile in data.keys():
|
|
print(profile)
|
|
if os.path.isfile("%s//%s.csv" % (out_csv_loc, profile)):
|
|
|
|
#if the file already exists, check if our new date is there (if so, overwrite)
|
|
file_check=pd.read_csv("%s//%s.csv" % (out_csv_loc, profile), dtype={'Date':str}) #
|
|
|
|
if str(survey_date) in [str(file_check['Date'][i]) for i in range(0, len(file_check['Date']))]:
|
|
|
|
file_check=file_check[file_check['Date']!=str(survey_date)]
|
|
file_check.to_csv("%s//%s.csv" % (out_csv_loc,profile), index=False)
|
|
|
|
#if the file exists, just append to the file
|
|
f=open("%s//%s.csv" % (out_csv_loc, profile), 'a')
|
|
else:
|
|
#otherwise make a new file
|
|
f=open("%s//%s.csv" % (out_csv_loc, profile), 'w')
|
|
f.writelines("Beach,Profile,Date,Easting,Northing,Chainage,Elevation\n")
|
|
|
|
|
|
for i in range(0, len(data[profile]['Easting'])):
|
|
if data[profile]['Elevation'][i]=="-":
|
|
z=""
|
|
else:
|
|
z=data[profile]['Elevation'][i]
|
|
|
|
f.writelines("%s,%s,%s,%s,%s,%s,%s\n" % (beach_name, profile,survey_date, data[profile]['Easting'][i],data[profile]['Northing'][i],data[profile]['Chainage'][i], z))
|
|
f.close()
|
|
|
|
return None
|
|
|
|
def plot_profiles(profile_info, profile, output_loc, LL_limit):
|
|
#plot the profile. expects output from CC_split_profile
|
|
|
|
YminorLocator=MultipleLocator(0.5)
|
|
XminorLocator=MultipleLocator(5)
|
|
|
|
|
|
fig,ax=plt.subplots(figsize=(8, 3))
|
|
|
|
num_plots=len(profile_info.keys())-1
|
|
colormap = plt.cm.jet
|
|
ax.set_prop_cycle(cycler('color', [colormap(i) for i in np.linspace(0, 0.9, num_plots)]))
|
|
|
|
max_y=0
|
|
for date in profile_info.keys():
|
|
if date!='info':
|
|
|
|
plt.plot(profile_info[date]['Chainage'], profile_info[date]['Elevation'], label=date)
|
|
try:
|
|
if max([i for i in profile_info[date]['Elevation'] if pd.isnull(i)==False])>max_y:
|
|
max_y=max([i for i in profile_info[date]['Elevation'] if pd.isnull(i)==False])
|
|
except:
|
|
print("empty elevation section for %s" % date)
|
|
|
|
|
|
plt.plot([LL_limit,LL_limit], [-1,max_y], 'r--', alpha=0.5, label="Landward Limit")
|
|
|
|
|
|
plt.xlabel('Chainage (m)',weight='bold')
|
|
plt.ylabel('Elevation (m AHD)',weight='bold')
|
|
plt.legend(loc='upper right', bbox_to_anchor=(1.3,1))
|
|
plt.title(profile)
|
|
plt.rcParams['font.size']=8
|
|
|
|
ax.set_ylim([-1,math.ceil(max_y)])
|
|
|
|
ax.xaxis.set_minor_locator(XminorLocator)
|
|
ax.yaxis.set_minor_locator(YminorLocator)
|
|
ax.xaxis.grid(True, which='minor', color='k', linestyle='-', alpha=0.3)
|
|
ax.yaxis.grid(True,which='minor',color='k', linestyle='-', alpha=0.3)
|
|
|
|
plt.grid(which='major', color='k', linestyle='-')
|
|
today=datetime.datetime.now().date().strftime('%Y%m%d')
|
|
plt.savefig(os.path.join(output_loc, '%s_%s.png' % (today, profile)),bbox_inches='tight',dpi=900)
|
|
plt.clf()
|
|
|
|
return None
|
|
|
|
|
|
def CC_split_profile(file2read):
|
|
# this reads the profile files and splits it into dates
|
|
|
|
file_master=pd.read_csv(file2read)
|
|
|
|
beach_original=file_master['Beach'].tolist()
|
|
profile_original=file_master['Profile'].tolist()
|
|
date_original=file_master['Date'].tolist()
|
|
chainage_original=file_master['Chainage'].tolist()
|
|
elevation_original=file_master['Elevation'].tolist()
|
|
easting_original=file_master['Easting'].tolist()
|
|
northing_original=file_master['Northing'].tolist()
|
|
|
|
data={}
|
|
i=0
|
|
|
|
#add info on the beach and profile number
|
|
data['info']={'Profile':profile_original[0], 'Beach':beach_original[0]}
|
|
date_now=date_original[0]
|
|
|
|
while i<len(file_master):
|
|
chainage_tmp=[]
|
|
elevation_tmp=[]
|
|
easting_tmp=[]
|
|
northing_tmp=[]
|
|
while i<len(file_master) and date_now==date_original[i]:
|
|
chainage_tmp.append(chainage_original[i])
|
|
elevation_tmp.append(elevation_original[i])
|
|
easting_tmp.append(easting_original[i])
|
|
northing_tmp.append(northing_original[i])
|
|
i=i+1
|
|
|
|
data[date_now]={'Beach': beach_original[i-1], 'Profile':profile_original[i-1],'Easting': easting_tmp, 'Northing':northing_tmp, 'Elevation':elevation_tmp, 'Chainage':chainage_tmp}
|
|
|
|
if i<len(file_master):
|
|
date_now=date_original[i]
|
|
|
|
return data
|
|
|
|
def profile_plots_volume(csv_loc, LL_xlsx, output_xlsx, graph_location):
|
|
#get a list of all csvs which will each be analysed
|
|
file_list=[]
|
|
for file in os.listdir(csv_loc):
|
|
if file.endswith(".csv"):
|
|
file_list.append(os.path.join(csv_loc, file))
|
|
|
|
#now read the LL file
|
|
LL_limit_file=pd.read_excel(LL_xlsx, 'profile_locations')
|
|
LL_info={}
|
|
for i in range(0, len(LL_limit_file)):
|
|
#make a dictionary that alllows you to search the LL
|
|
prof="%s_%s" % (LL_limit_file['Profile'][i].split(" ")[0], LL_limit_file['Profile'][i].split(" ")[-1])
|
|
|
|
LL_info[prof]=LL_limit_file['Landward Limit'][i]
|
|
|
|
all_dates=[]
|
|
results_volume={}
|
|
for file in file_list:
|
|
#read the profile data - this should have all dates
|
|
profile_data=CC_split_profile(file)
|
|
profile=profile_data['info']['Profile']
|
|
|
|
#plot all of the profiles
|
|
print(profile)
|
|
plot_profiles(profile_data, profile, graph_location,LL_info[profile])
|
|
|
|
results_volume[profile]={}
|
|
|
|
#nowgo through each date and do a neilson volume calculations
|
|
for date in profile_data.keys():
|
|
if date!='info':
|
|
if date not in all_dates:
|
|
all_dates.append(date)
|
|
|
|
chainage=profile_data[date]['Chainage']
|
|
elevation=[0 if pd.isnull(profile_data[date]['Elevation'][i]) else profile_data[date]['Elevation'][i] for i in range(0, len(profile_data[date]['Elevation']))]
|
|
LL_limit=LL_info[profile]
|
|
#do a neilson calculation to get the ZSA volume
|
|
if len(elevation)>2:
|
|
#if there aren't enough available points don't do it
|
|
volume=neilson_volumes.volume_available(chainage, elevation, LL_limit)
|
|
if volume<0:
|
|
volume=0
|
|
print('%s %s has a negative volume available' % (profile, date))
|
|
else:
|
|
volume=0
|
|
|
|
results_volume[profile][date]=volume
|
|
|
|
#write an excel sheet which summarises the data
|
|
workbook = xlsxwriter.Workbook(output_xlsx)
|
|
worksheet=workbook.add_worksheet('Volumes')
|
|
|
|
row=0
|
|
col=0
|
|
|
|
worksheet.write(row, col, 'Profile')
|
|
for date in all_dates:
|
|
col=col+1
|
|
worksheet.write(row, col, date)
|
|
|
|
col=0
|
|
row=1
|
|
|
|
for prof in results_volume.keys():
|
|
|
|
worksheet.write(row, col, prof)
|
|
|
|
for date in all_dates:
|
|
col=col+1
|
|
try:
|
|
vol=results_volume[prof][date]
|
|
except KeyError:
|
|
print("error with profile %s on %s" % (prof, date))
|
|
vol=None
|
|
worksheet.write(row, col, vol)
|
|
col=0
|
|
row=row+1
|
|
return results_volume
|
|
|
|
def remove_temp_files(directory):
|
|
for f in os.listdir(directory):
|
|
os.unlink(os.path.join(directory, f))
|
|
|
|
return None
|
|
|
|
###############################################################################
|
|
########################### RUN CODE ##########################################
|
|
params_file=pd.read_excel(input_file, sheet_name="PARAMS")
|
|
|
|
for i in range(0, len(params_file)): #0, len(params_file)
|
|
print("Starting to process %s" % params_file['Beach'][i])
|
|
beach=params_file['Beach'][i]
|
|
survey_date=params_file['SURVEY DATE'][i]
|
|
input_las5=params_file['INPUT LAS4'][i]
|
|
crop_swash_poly=params_file['CROP SWASH POLY'][i]
|
|
heatmap_crop_poly=params_file['HEATMAP CROP POLY'][i]
|
|
final_las = params_file['FINAL LAS'][i]
|
|
heatmap_las = params_file['HEATMAP LAS'][i]
|
|
zone_MGA=params_file['ZONE MGA'][i]
|
|
output_poly_name=params_file['OUTPUT POLY NAME'][i]
|
|
path_2_output_poly=params_file['PATH TO OUTPUT'][i]
|
|
output_raster=params_file['OUTPUT RASTER'][i]
|
|
input_csv=params_file['INPUT CSV'][i]
|
|
tmp_csv = params_file['TMP CSV'][i]
|
|
LL_file=params_file['LL FILE'][i]
|
|
csv_loc=params_file['OUT CSV LOC'][i]
|
|
graph_loc = params_file['GRAPH LOC'][i]
|
|
volume_output=params_file['VOLUME OUTPUT'][i]
|
|
tmp_dir=params_file['TEMP DIR'][i]
|
|
int_dir=params_file['INTERIM DIR'][i]
|
|
|
|
# crop and get the output las
|
|
crop_las(input_las5, crop_swash_poly, final_las, path_2_lastools)
|
|
|
|
#now crop out the heatmap las
|
|
crop_las(final_las, heatmap_crop_poly, heatmap_las, path_2_lastools)
|
|
|
|
#create a polygon to crop a raster
|
|
las_boundary(heatmap_las, output_poly_name, path_2_output_poly, path_2_lastools, zone_MGA)
|
|
#make a raster
|
|
make_raster(heatmap_las, output_raster, path_2_lastools, keep_only_ground=True)
|
|
|
|
#extract the points and get volumes
|
|
extract_pts(final_las, input_csv, tmp_csv, path_2_lastools, keep_only_ground=True)
|
|
process_tmp_csv(tmp_csv, survey_date, csv_loc, beach)
|
|
|
|
#colourise the point cloud
|
|
|
|
|
|
#delete the temp files from the tmp_dir and the interim_dir
|
|
remove_temp_files(tmp_dir)
|
|
#remove_temp_files(int_dir)
|
|
|
|
print("doing the volume analysis")
|
|
test=profile_plots_volume(csv_loc, LL_file, volume_output, graph_loc)
|