def __init__(self):
print('init LI_IMX377_MIPI_M12')
self.width = CAMERA_WIDTH
self.height = CAMERA_HEIGHT
self.cap = None
self.mtx, self.dist = load_calibration()
self.open()
def load_and_prepare_img(src_img_filename):
src_img = cv2.imread(src_img_filename)
camera_matrix, dist_coefs = load_calibration()
undistorted = undistort_image(camera_matrix, dist_coefs, src_img)
height, width, color_depth = undistorted.shape
clipped = undistorted[10:height-10, 100:width-100]
grey = cv2.cvtColor(clipped, cv2.COLOR_BGR2GRAY)
return {
'original_img': src_img,
'grey_img': grey,
'filename': src_img_filename,
'basename': basename(src_img_filename)
}
M_inv = cv2.getPerspectiveTransform(dst, src)
# Only for creating the final video visualization
X_b = [574, 706, 706, 574]
Y_b = [719, 719, 0, 0]
src_ = np.floor(np.float32([[x[0], y[0]], [x[1], y[1]],[x[2], y[2]], [x[3], y[3]]]) / (input_scale*2))
dst_ = np.floor(np.float32([[X_b[0], Y_b[0]], [X_b[1], Y_b[1]],[X_b[2], Y_b[2]], [X_b[3], Y_b[3]]]) / (input_scale*2))
M_b = cv2.getPerspectiveTransform(src_, dst_)
# Only for creating the final video visualization
# Threshold for color and gradient thresholding
s_thresh, sx_thresh, dir_thresh, m_thresh, r_thresh = (120, 255), (20, 100), (0.7, 1.3), (30, 100), (200, 255)
# load the calibration
calib_file = 'camera_cal/calibration_pickle.p'
mtx, dist = load_calibration(calib_file)
def abs_sobel_thresh(img, orient='x', sobel_kernel=3, thresh=(0, 255)):
# Apply the following steps to img
# 1) Convert to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
# 2) Take the derivative in x or y given orient = 'x' or 'y'
# 3) Take the absolute value of the derivative or gradient
if orient == 'x':
abs_sobel = np.absolute(cv2.Sobel(gray, cv2.CV_64F, 1, 0, ksize=sobel_kernel))
if orient == 'y':
abs_sobel = np.absolute(cv2.Sobel(gray, cv2.CV_64F, 0, 1, ksize=sobel_kernel))
def process_vanadium_data(vanadium,
empty_instr,
lambda_binning,
calibration=None,
**absorp):
"""
Create corrected vanadium dataset
Correction applied to Vanadium data only
1. Subtract empty instrument
2. Correct absorption
3. Use calibration for grouping
4. Focus into groups
Parameters
----------
vanadium : Vanadium nexus datafile
empty_instr: Empty instrument nexus file
lambda_binning: format=(lambda_min, lambda_min, number_of_bins)
lambda_min and lambda_max are in Angstroms
calibration: calibration file
Mantid format
**absorp: dictionary containing information to correct absorption for
sample and vanadium
Only the inputs related to Vanadium will be selected to calculate
the correction
see docstrings of powder_reduction for more details
see help of Mantid's algorithm CylinderAbsorption for details
https://docs.mantidproject.org/nightly/algorithms/CylinderAbsorption-v1.html
"""
vana_red = process_event_data(vanadium, lambda_binning)
ec_red = process_event_data(empty_instr, lambda_binning)
# remove 'spectrum' from wavelength coordinate and match this coordinate
# between Vanadium and Empty instrument data
min_lambda = vana_red.coords['wavelength'].values[:, 0].min()
max_lambda = vana_red.coords['wavelength'].values[:, 1].max()
vana_red.coords['wavelength'] = sc.Variable(['wavelength'],
unit=sc.units.angstrom,
values=np.linspace(min_lambda,
max_lambda,
num=2))
ec_red.coords['wavelength'] = sc.Variable(['wavelength'],
unit=sc.units.angstrom,
values=np.linspace(min_lambda,
max_lambda,
num=2))
# vana - EC
ec_red.coords['wavelength'] = vana_red.coords['wavelength']
vana_red.bins.concatenate(-ec_red, out=vana_red)
del ec_red
# Absorption correction applied
if bool(absorp):
# The values of number_density, scattering and attenuation are
# hard-coded since they must correspond to Vanadium. Only radius and
# height of the Vanadium cylindrical sample shape can be set. The
# names of these inputs if present have to be renamed to match
# the requirements of Mantid's algorithm CylinderAbsorption
# Create dictionary to calculate absorption correction for Vanadium.
absorp_vana = {
key.replace('Vanadium', 'Sample'): value
for key, value in absorp.items() if 'Vanadium' in key
}
absorp_vana['SampleNumberDensity'] = 0.07118
absorp_vana['ScatteringXSection'] = 5.16
absorp_vana['AttenuationXSection'] = 4.8756
correction = absorption_correction(vanadium, lambda_binning,
**absorp_vana)
# the 3 following lines of code are to place info about source and
# sample position at the right place in the correction dataArray in
# order to proceed to the normalization
del correction.coords['source_position']
del correction.coords['sample_position']
del correction.coords['position']
lambda_min, lambda_max, number_bins = lambda_binning
edges_lambda = sc.Variable(['wavelength'],
unit=sc.units.angstrom,
values=np.linspace(lambda_min,
lambda_max,
num=number_bins))
correction = sc.rebin(correction, 'wavelength', edges_lambda)
vana_red = vana_red.bins / sc.lookup(func=correction, dim='wavelength')
del correction
# Calibration
# Load
input_load_cal = {'InstrumentFilename': 'WISH_Definition.xml'}
calvana = load_calibration(calibration, mantid_args=input_load_cal)
# Merge table with detector->spectrum mapping from vanadium
# (implicitly checking that detectors between vanadium and
# calibration are the same)
cal_vana = sc.merge(calvana, vana_red.coords['detector_info'].value)
del calvana
# Compute spectrum mask from detector mask
maskvana = sc.groupby(cal_vana['mask'], group='spectrum').any('detector')
# Compute spectrum groups from detector groups
gvana = sc.groupby(cal_vana['group'], group='spectrum')
groupvana = gvana.min('detector')
assert sc.identical(groupvana, gvana.max('detector')), \
("Calibration table has mismatching group "
"for detectors in same spectrum")
vana_red.coords['group'] = groupvana.data
vana_red.masks['mask'] = maskvana.data
# convert to d-spacing with calibration
vana_dspacing = convert_with_calibration(vana_red, cal_vana)
del vana_red, cal_vana
# Focus
focused_vana = \
sc.groupby(vana_dspacing, group='group').bins.concatenate('spectrum')
del vana_dspacing
return focused_vana
def powder_reduction(sample='sample.nxs',
calibration=None,
vanadium=None,
empty_instr=None,
lambda_binning=(0.7, 10.35, 5615),
**absorp):
"""
Simple WISH reduction workflow
Note
----
The sample data were not recorded using the same layout
of WISH as the Vanadium and empty instrument. That's why:
- loading calibration for Vanadium used a different IDF
- the Vanadium correction involved cropping the sample data
to the first 5 groups (panels)
----
Corrections applied:
- Vanadium correction
- Absorption correction
- Normalization by monitors
- Conversion considering calibration
- Masking and grouping detectors into panels
Parameters
----------
sample: Nexus event file
calibration: .cal file following Mantid's standards
The columns correspond to detectors' IDs, offset, selection of
detectors and groups
vanadium: Nexus event file
empty_instr: Nexus event file
lambda_binning: min, max and number of steps for binning in wavelength
min and max are in Angstroms
**absorp: dictionary containing information to correct absorption for
Sample and Vanadium.
There could be only up to two elements related to the correction
for Vanadium:
the radius and height of the cylindrical sample shape.
To distinguish them from the inputs related to the sample, their
names in the dictionary are 'CylinderVanadiumRadius' and
'CylinderVanadiumHeight'. The other keysof the 'absorp'
dictionary follow Mantid's syntax and are related to the sample
data only.
See help of Mantid's algorithm CylinderAbsorption for details
https://docs.mantidproject.org/nightly/algorithms/CylinderAbsorption-v1.html
Returns
-------
Scipp dataset containing reduced data in d-spacing
Hints
-----
To plot the output data, one can histogram in d-spacing and sum according
to groups using scipp.histogram and sc.sum, respectively.
"""
# Load counts
sample_data = scn.load(sample,
advanced_geometry=True,
load_pulse_times=False,
mantid_args={'LoadMonitors': True})
# Load calibration
if calibration is not None:
input_load_cal = {"InstrumentName": "WISH"}
cal = load_calibration(calibration, mantid_args=input_load_cal)
# Merge table with detector->spectrum mapping from sample
# (implicitly checking that detectors between sample and calibration
# are the same)
cal_sample = sc.merge(cal, sample_data.coords['detector_info'].value)
# Compute spectrum mask from detector mask
mask = sc.groupby(cal_sample['mask'], group='spectrum').any('detector')
# Compute spectrum groups from detector groups
g = sc.groupby(cal_sample['group'], group='spectrum')
group = g.min('detector')
assert sc.identical(group, g.max('detector')), \
"Calibration table has mismatching group for detectors in same spectrum"
sample_data.coords['group'] = group.data
sample_data.masks['mask'] = mask.data
del cal
# Correct 4th monitor spectrum
# There are 5 monitors for WISH. Only one, the fourth one, is selected for
# correction (like in the real WISH workflow).
# Select fourth monitor and convert from tof to wavelength
mon4_lambda = scn.convert(sample_data.attrs['monitor4'].values,
'tof',
'wavelength',
scatter=False)
# Spline background
# mon4_spline_background = bspline_background(mon4_lambda,
# sc.Dim('wavelength'),
# smoothing_factor=70)
# Smooth monitor
mon4_smooth = smooth_data(
mon4_lambda, # mon4_spline_background,
dim='wavelength',
NPoints=40)
# Delete intermediate data
del mon4_lambda # , mon4_spline_background
# Correct data
# 1. Normalize to monitor
# Convert to wavelength (counts)
sample_lambda = scn.convert(sample_data, 'tof', 'wavelength', scatter=True)
del sample_data
# Rebin monitors' data
lambda_min, lambda_max, number_bins = lambda_binning
edges_lambda = sc.Variable(['wavelength'],
unit=sc.units.angstrom,
values=np.linspace(lambda_min,
lambda_max,
num=number_bins))
mon_rebin = sc.rebin(mon4_smooth, 'wavelength', edges_lambda)
# Realign sample data
sample_lambda = sample_lambda.bins / sc.lookup(func=mon_rebin,
dim='wavelength')
del mon_rebin, mon4_smooth
# 2. absorption correction
if bool(absorp):
# Copy dictionary of absorption parameters
absorp_sample = absorp.copy()
# Remove input related to Vanadium if present in absorp dictionary
found_vana_info = [
key for key in absorp_sample.keys() if 'Vanadium' in key
]
for item in found_vana_info:
absorp_sample.pop(item, None)
# Calculate absorption correction for sample data
correction = absorption_correction(sample, lambda_binning,
**absorp_sample)
# the 3 following lines of code are to place info about source and
# sample position at the right place in the correction dataArray in
# order to proceed to the normalization
del correction.coords['source_position']
del correction.coords['sample_position']
del correction.coords['position']
correction_rebin = sc.rebin(correction, 'wavelength', edges_lambda)
del correction
sample_lambda = sample_lambda.bins / sc.lookup(func=correction_rebin,
dim='wavelength')
# 3. Convert to d-spacing with option of taking calibration into account
if calibration is None:
# No calibration data, use standard convert algorithm
sample_dspacing = scn.convert(sample_lambda, 'tof', 'dspacing')
else:
# Calculate dspacing from calibration file
sample_dspacing = convert_with_calibration(sample_lambda, cal_sample)
del cal_sample
del sample_lambda
# 4. Focus panels
# Assuming sample is in d-spacing: Focus into groups
focused = sc.groupby(sample_dspacing,
group='group').bins.concatenate('spectrum')
del sample_dspacing
# 5. Vanadium correction (requires Vanadium and Empty instrument)
if vanadium is not None and empty_instr is not None:
print("Proceed with reduction of Vanadium data ")
vana_red_focused = process_vanadium_data(vanadium, empty_instr,
lambda_binning, calibration,
**absorp)
# The following selection of groups depends on the loaded data for
# Sample, Vanadium and Empty instrument
focused = focused['group', 0:5].copy()
# histogram vanadium for normalizing + cleaning 'metadata'
d_min, d_max, number_dbins = (1., 10., 2000)
edges_dspacing = sc.Variable(['dspacing'],
unit=sc.units.angstrom,
values=np.linspace(d_min,
d_max,
num=number_dbins))
vana_histo = sc.histogram(vana_red_focused, edges_dspacing)
del vana_red_focused
vana_histo.coords['detector_info'] = focused.coords[
'detector_info'].copy()
# del vana_histo.coords['source_position']
# del vana_histo.coords['sample_position']
# normalize by vanadium
result = focused.bins / sc.lookup(func=vana_histo, dim='dspacing')
del vana_histo, focused
return result
from moviepy.editor import VideoFileClip
from Frame import Frame
from Line import Line
import cv2
import numpy as np
import matplotlib.image as mpimg
import matplotlib.pyplot as plt
from Line import LaneSide
from calibration import load_calibration, undistort
# load calibration
mtx, dist = load_calibration()
l_line = Line(LaneSide.Left)
r_line = Line(LaneSide.Right)
# image = mpimg.imread('test_images/straight_lines1.jpg')
#image = mpimg.imread('test_images/straight_lines2.jpg')
# image = mpimg.imread('test_images/test1.jpg')
#image = mpimg.imread('test_images/test2.jpg')
image = mpimg.imread('test_images/challenge.jpg')
# image = mpimg.imread('test_images/challenge2.png')
# image = mpimg.imread('test_images/challenge4.jpg')
# image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
frames = []
M_inv = cv2.getPerspectiveTransform(dst, src)
# Only for creating the final video visualization
X_b = [574, 706, 706, 574]
Y_b = [719, 719, 0, 0]
src_ = np.floor(np.float32([[x[0], y[0]], [x[1], y[1]],[x[2], y[2]], [x[3], y[3]]]) / (input_scale*2))
dst_ = np.floor(np.float32([[X_b[0], Y_b[0]], [X_b[1], Y_b[1]],[X_b[2], Y_b[2]], [X_b[3], Y_b[3]]]) / (input_scale*2))
M_b = cv2.getPerspectiveTransform(src_, dst_)
# Only for creating the final video visualization
# Threshold for color and gradient thresholding
s_thresh, sx_thresh, dir_thresh, m_thresh, r_thresh = (120, 255), (20, 100), (0.7, 1.3), (30, 100), (200, 255)
# load the calibration
calib_file = 'calibration_pickle.p'
mtx, dist = load_calibration(calib_file)
def abs_sobel_thresh(img, orient='x', sobel_kernel=3, thresh=(0, 255)):
# Apply the following steps to img
# 1) Convert to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
# 2) Take the derivative in x or y given orient = 'x' or 'y'
# 3) Take the absolute value of the derivative or gradient
if orient == 'x':
abs_sobel = np.absolute(cv2.Sobel(gray, cv2.CV_64F, 1, 0, ksize=sobel_kernel))
if orient == 'y':
abs_sobel = np.absolute(cv2.Sobel(gray, cv2.CV_64F, 0, 1, ksize=sobel_kernel))
