import numpy as np
import cv2
import datetime
def split_xy(xy_s):
x, y = list(), list()
for xy in xy_s:
x.append(xy[0])
y.append(xy[1])
x = np.array(x)
y = np.array(y)
return x, y
def reg1dim(x, y):
n = len(x)
a = ((np.dot(x, y) - y.sum() * x.sum()/n) / ((x ** 2).sum() - x.sum()**2 / n))
b = (y.sum() - a * x.sum()) / n
return a, b
def plot_line(im0, a, b):
imsize = im0.shape[0]
if 0 <= int(-a/b) <= imsize:
cv2.line(im0, (int(-b/a), 0), (int((imsize - b) / a), imsize), (112, 164, 184), 2, lineType=cv2.LINE_AA)
else:
cv2.line(im0, (0, int(b)), (imsize, int(a * imsize + b)), (112, 164, 184), 2, lineType=cv2.LINE_AA)
def plot_part_line(im0, center_points):
# TODO:改善の余地あり
for i in range(len(center_points) - 1):
cv2.line(im0, center_points[i], center_points[i + 1], (112, 164, 184), 1, lineType=cv2.LINE_AA)
def calc_mm_per_pixel(center_xys, real_wl_interval=50.):
mm_per_pixel_sum = 0.0
for i in range(len(center_xys) - 1):
mm_per_pixel_sum += (50. / np.linalg.norm(np.array(center_xys[i + 1]) - np.array(center_xys[i]), ord=2))
mm_per_pixel = (mm_per_pixel_sum / (len(center_xys) - 1))
return mm_per_pixel
def calc_match_list(peak_xy, prev_peak_xy, thresh_distance=30):
match_list = []
if len(prev_peak_xy) <= len(peak_xy):
for prev_id, prev_xy in enumerate(prev_peak_xy):
prev_xy = np.array(prev_xy)
match_cur_id = -1
min_distance = float('inf')
for cur_id, cur_xy in enumerate(peak_xy):
cur_xy = np.array(cur_xy)
xy_distance = np.sum(np.abs(cur_xy - prev_xy))
if xy_distance < min_distance:
min_distance = xy_distance
match_cur_id = cur_id
if min_distance < thresh_distance:
match_list.append([prev_xy, peak_xy[match_cur_id]])
else:
pass
else:
for cur_id, cur_xy in enumerate(peak_xy):
cur_xy = np.array(cur_xy)
match_prev_id = -1
min_distance = float('inf')
for prev_id, prev_xy in enumerate(prev_peak_xy):
prev_xy = np.array(prev_xy)
xy_distance = np.sum(np.abs(cur_xy - prev_xy))
if xy_distance < min_distance:
min_distance = xy_distance
match_prev_id = prev_id
if min_distance < thresh_distance:
match_list.append([prev_peak_xy[match_prev_id], cur_xy])
else:
pass
return match_list
def calc_direction_vec(center_points):
x_diff_list = list()
y_diff_list = list()
for i in range(len(center_points) - 1):
x_diff_list.append(center_points[i + 1][0] - center_points[i][0])
y_diff_list.append(center_points[i + 1][1] - center_points[i][1])
scope_dir_vec = np.array([sum(x_diff_list)/len(x_diff_list), sum(y_diff_list)/len(y_diff_list)])
l2 = np.linalg.norm(scope_dir_vec, ord=2, axis=-1, keepdims=True)
l2[l2 == 0] = 1
return scope_dir_vec/l2
def sort_center_points(center_points):
x_s, y_s = split_xy(center_points)
x_diverse, y_diverse = max(x_s) - min(x_s), max(y_s) - min(y_s)
# diverseが大きい方に準拠してsortする
sorted_center = sorted(center_points, key=lambda xy: xy[0]) if y_diverse <= x_diverse else \
sorted(center_points, key=lambda xy: xy[1])
return sorted_center
def calc_valid_length(match_list, scope_dir_vec, mm_per_pixel):
valid_mm_length = list()
for match_xy in match_list:
move_vec = match_xy[1] - match_xy[0]
valid_pixel_length = np.dot(move_vec, scope_dir_vec)
valid_mm_length.append(valid_pixel_length * mm_per_pixel)
valid_mm_length = sum(valid_mm_length) / len(valid_mm_length)
return valid_mm_length
def show_formula_mat(formula):
formula_mat = np.zeros((50, 256, 3), dtype=np.uint8)
cv2.putText(formula_mat, '{:7.1f} mm'.format(formula), (10, 30), cv2.FONT_HERSHEY_DUPLEX,
1, (0, 200, 200), thickness=3, lineType=cv2.LINE_AA)
cv2.imshow("formula", formula_mat)
return formula_mat
def eval_linear_score(cur_center_points, thresh_val=0.9):
if 3 <= len(cur_center_points):
base_point = np.array(cur_center_points[0])
inter_vec_list = []
for point in cur_center_points[1:]:
point = np.array(point)
relative_vec = point - base_point
inter_vec = relative_vec / np.linalg.norm(relative_vec)
inter_vec_list.append(inter_vec)
base_vec = inter_vec_list[0]
linear_score = 1.0
for vec in inter_vec_list[1:]:
dot = np.dot(base_vec, vec)
linear_score *= dot
base_vec = vec
if thresh_val < linear_score:
return True
else:
print(f"{datetime.datetime.now()} linear_score is low")
return False
else:
return True
def eval_interval_score(cur_center_points, thresh_val=0.9):
if 3 <= len(cur_center_points):
base_point = np.array(cur_center_points[0])
relative_vec_list = []
relative_norm_list = []
for point in cur_center_points[1:]:
point = np.array(point)
relative_vec = point - base_point
relative_vec_list.append(relative_vec)
relative_norm_list.append(np.linalg.norm(relative_vec))
base_point = point
base_vec = relative_vec_list[0]
interval_score = 1.0
for vec in relative_vec_list[1:]:
vec_diff = (vec - base_vec) / max(relative_norm_list)
interval_score *= (1.0 - np.linalg.norm(vec_diff))
base_vec = vec
if thresh_val < interval_score:
return True
else:
print(f"{datetime.datetime.now()} interval_score is low")
return False
else:
return True
def eval_series_score(mpp, p_mpp, thresh_val=0.9):
if p_mpp is not None:
series_score = 1.0 - abs((mpp - p_mpp) / max(mpp, p_mpp))
if thresh_val < series_score:
return True
else:
print("series_score is low")
return False
else:
return True
