def extract_table(table_body, __line__, lines=None):
# Deciding variable
if (__line__ == 1):
# Check if table image is bordered or borderless
temp_lines_hor, temp_lines_ver = line_detection(table_body)
else:
temp_lines_hor, temp_lines_ver = lines
if len(temp_lines_hor) == 0 or len(temp_lines_ver) == 0:
print("Either Horizontal Or Vertical Lines Not Detected")
return None
table = table_body.copy()
x = 0
y = 0
k = 0
points = []
print("[Table status] : Processing table with lines")
# Remove same lines detected closer
for x1, y1, x2, y2 in temp_lines_ver:
point = []
for x3, y3, x4, y4 in temp_lines_hor:
try:
k += 1
x, y = line_intersection(x1, y1, x2, y2, x3, y3, x4, y4)
point.append([x, y])
except:
continue
points.append(point)
for point in points:
for x, y in point:
cv2.line(table, (x, y), (x, y), (0, 0, 255), 8)
cv2.imshow("intersection", table)
cv2.waitKey(0)
# boxno = -1
box = []
flag = 1
lastCache = []
## creating bounding boxes of cells from the points detected
## This is still under work and might fail on some images
for i, row in enumerate(points):
limitj = len(row)
currentVala = []
for j, col in enumerate(row):
if (j == limitj - 1):
break
if (i == 0):
nextcol = row[j + 1]
lastCache.append([
col[0], col[1], nextcol[0], nextcol[1], 9999, 9999, 9999,
9999
])
else:
nextcol = row[j + 1]
currentVala.append([
col[0], col[1], nextcol[0], nextcol[1], 9999, 9999, 9999,
9999
])
# Matching
flag = 1
index = []
for k, last in enumerate(lastCache):
if (col[1] == last[1]) and lastCache[k][4] == 9999:
lastCache[k][4] = col[0]
lastCache[k][5] = col[1]
if lastCache[k][4] != 9999 and lastCache[k][6] != 9999:
box.append(lastCache[k])
index.append(k)
flag = 1
if (nextcol[1] == last[3]) and lastCache[k][6] == 9999:
lastCache[k][6] = nextcol[0]
lastCache[k][7] = nextcol[1]
if lastCache[k][4] != 9999 and lastCache[k][6] != 9999:
box.append(lastCache[k])
index.append(k)
flag = 1
if len(lastCache) != 0:
if lastCache[k][4] == 9999 or lastCache[k][6] == 9999:
flag = 0
# print(index)
for k in index:
lastCache.pop(k)
# tranfsering
if flag == 0:
for last in lastCache:
if last[4] == 9999 or last[6] == 9999:
currentVala.append(last)
if (i != 0):
lastCache = currentVala
## Visualizing the cells ##
# count = 1
# for i in box:
# cv2.rectangle(table_body, (i[0], i[1]), (i[6], i[7]), (int(i[7]%255),0,int(i[0]%255)), 2)
# # count+=1
# cv2.imshow("cells",table_body)
# cv2.waitKey(0)
############################
return box
import numpy as np
from Functions.line_detection import line_detection
base_path = '/Volumes/my_disk/company/sensedeal/217_PycharmProject/bbtv/PaddleOCR-1.0-2021/doc/my_imgs_11/'
image_name_list = os.listdir(base_path)
for image_name in image_name_list:
image = cv2.imread(base_path + image_name)
img_h, img_w, _ = np.shape(image)
if img_h > 1000:
resize_h = 1000
else:
resize_h = img_w
temp_lines_hor, temp_lines_ver = line_detection(image)
temp_lines_hor.append([0, 0, img_w, 0])
temp_lines_hor.append([0, img_h, img_w, img_h])
temp_lines_ver.append([0, 0, 0, img_h])
temp_lines_ver.append([img_w, 0, img_w, img_h])
show_img = image.copy()
temp = []
for line in temp_lines_hor:
x1, y1, x2, y2 = line
if abs(x1 - x2) / img_w > 0.618:
temp.append([x1, y1, x2, y2])
temp_lines_hor = temp
for i in temp_lines_hor:
cv2.line(show_img, (i[0], i[1]), (i[2], i[3]), (0, 0, 255), 2)