156 lines
3.4 KiB
Python
156 lines
3.4 KiB
Python
import numpy as np
|
|
from scipy.optimize import curve_fit
|
|
import utils
|
|
import sys
|
|
|
|
data = utils.load( sys.argv[1] )
|
|
|
|
"""
|
|
find fill points
|
|
"""
|
|
points = []
|
|
|
|
points += utils.find_point( data[ : , 0 ] , 0 ) # x_min
|
|
points += utils.find_point(
|
|
data[ : , data.shape[1] - 1 ],
|
|
data.shape[1] - 1 ,
|
|
) # x_max
|
|
|
|
index_min = 0
|
|
while data.shape[0] - 1 > index_min:
|
|
index_min += 1
|
|
if len( utils.find_point( data[ index_min , : ] , index_min , 'y' ) ) == 3:
|
|
break
|
|
|
|
points.append(
|
|
utils.find_point( data[ index_min , : ] , index_min , 'y' )[1]
|
|
) # y_min
|
|
|
|
index_max = data.shape[0] - 1
|
|
while index_min < index_max:
|
|
index_max -= 1
|
|
if len( utils.find_point( data[ index_max , : ] , index_max , 'y' ) ) == 3:
|
|
break
|
|
|
|
|
|
points.append(
|
|
utils.find_point( data[ index_max , : ] , index_max , 'y' )[1]
|
|
) # y_max
|
|
|
|
small_data = utils.compress( data , 5 )
|
|
|
|
points = utils.big_to_small( points , 5 )
|
|
|
|
# size - 1
|
|
points[ 1 ][ 1 ] -= 1
|
|
points[ 3 ][ 0 ] -= 1
|
|
|
|
# little shift to be inside the light
|
|
points[ 2 ][ 1 ] += 3
|
|
points[ 3 ][ 1 ] += 3
|
|
|
|
"""
|
|
fill data
|
|
"""
|
|
|
|
extremum = []
|
|
for point in points:
|
|
taken_points = utils.small_to_big(
|
|
utils.fill( small_data , point , 1000 ),
|
|
5
|
|
)
|
|
extremum.append( [
|
|
np.min( taken_points[ : , 1 ] ),
|
|
np.max( taken_points[ : , 1 ] ),
|
|
np.min( taken_points[ : , 0 ] ),
|
|
np.max( taken_points[ : , 0 ] ),
|
|
] )
|
|
|
|
border = {
|
|
'x': {
|
|
'min': extremum[0][1] + 1,
|
|
'max': extremum[1][0] ,
|
|
},
|
|
'y': {
|
|
'min': extremum[2][3] + 1,
|
|
'max': extremum[3][2] ,
|
|
},
|
|
}
|
|
|
|
"""
|
|
label deletion
|
|
"""
|
|
|
|
mean_data = np.mean( data[
|
|
border['y']['min'] : border['y']['max'],
|
|
border['x']['min'] : border['x']['max']
|
|
] , axis = 0 )
|
|
|
|
gauss = lambda x , sigma , mu , a , b : a * (
|
|
1 / sigma * np.sqrt(
|
|
2 * np.pi
|
|
) * np.exp(
|
|
- ( x - mu ) ** 2 / ( 2 * sigma ** 2 )
|
|
)
|
|
) + b
|
|
abciss = np.arange(
|
|
border['x']['min'],
|
|
border['x']['max'],
|
|
1
|
|
)
|
|
guess_params = [
|
|
1 ,
|
|
border['x']['min'] + ( border['x']['max'] - border['x']['min'] ) // 2,
|
|
np.max( mean_data ) ,
|
|
np.min( mean_data ) ,
|
|
]
|
|
|
|
first_estimate = curve_fit(
|
|
gauss ,
|
|
abciss ,
|
|
mean_data ,
|
|
guess_params
|
|
)[0]
|
|
|
|
part_data = [
|
|
mean_data[ : mean_data.shape[0] // 2 ],
|
|
mean_data[ mean_data.shape[0] // 2 : ]
|
|
]
|
|
part_abciss = [
|
|
abciss[ : abciss.shape[0] // 2 ],
|
|
abciss[ abciss.shape[0] // 2 : ]
|
|
]
|
|
part_result = []
|
|
for i in range( 2 ):
|
|
part_result.append(
|
|
curve_fit(
|
|
gauss ,
|
|
part_abciss[i],
|
|
part_data[i] ,
|
|
first_estimate
|
|
)
|
|
)
|
|
|
|
cov = np.array( [
|
|
np.sum( np.diag( part_result[i][1] ) ) for i in range( 2 )
|
|
] )
|
|
i = np.argmax( cov ) # part where the label is
|
|
|
|
derivee = np.convolve(
|
|
np.gradient( part_data[i] ),
|
|
np.ones( part_data[i].shape[0] // 100 ),
|
|
'same',
|
|
)
|
|
start_label = np.argmax( derivee )
|
|
end_label = np.argmin( derivee[ start_label :: ( - 1 ) ** i ] )
|
|
|
|
keys = [ 'min' , 'max' ]
|
|
|
|
border['x'][keys[i]] += ( - 1 ) ** i * ( start_label + end_label )
|
|
|
|
import matplotlib.pyplot as plt
|
|
plt.imshow( data[
|
|
border[ 'y' ][ 'min' ] : border[ 'y' ][ 'max' ],
|
|
border[ 'x' ][ 'min' ] : border[ 'x' ][ 'max' ]
|
|
] )
|
|
plt.savefig( 'asset/test.png' )
|