0

我正在尝试使用 Python OpenCV 和 Hough 变换算法检测表格行并从图像中提取完整表格。我需要拥有每条线的所有坐标，目的是以相同的比例绘制同一张桌子。我了解霍夫变换的工作原理并尝试在没有 OpenCV 的情况下实现它，但在大图像上速度非常慢。

这是来自示例 OpenCV Hough Transfrom 的代码

import cv2
import numpy as np

img = cv2.imread('image1.jpg')
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
edges = cv2.Canny(gray, 50, 150, apertureSize=3)
cv2.imshow("image", edges)
cv2.waitKey(0)
minLineLength = 100
maxLineGap = 10
lines = cv2.HoughLinesP(edges, 1, np.pi / 180, 50, minLineLength, maxLineGap)
for line in lines:
    for x1, y1, x2, y2 in line:
        cv2.line(img, (x1, y1), (x2, y2), (0, 255, 0), 2)

cv2.imwrite('houghlines5.jpg', img)

Canny 边缘检测返回 Canny 边缘检测的图像Resulf

但是检测的结果是Housh Transform 的 Resulf

我不知道为什么霍夫变换会留下一些表格。你能推荐一些事情吗？也许另一种从图像中提取表格的方法？谢谢！

UPD。需要检测的原始图像原始表

4

2 回答 2

3

我认为你必须在执行之前准备好图像HoughLinesP()，例如去掉文本、使用参数或扩大边缘等。但是如果你想提取没有文本的模板，你可以通过绘图来让你的生活更轻松从此表中删除所有文本（通过搜索小轮廓并在轮廓上制作白色覆盖边界框）。在示例代码中，我做了三个步骤：一是制作没有 Hough 的模板，二是制作模板，三是HoughLines()制作HoughLinesP(). 希望它有点帮助。干杯!

例子：

import cv2
import numpy as np

### MAKING TEMPLATE WITHOUT HOUGH

# Read the image and make a copy then transform it to gray colorspace,
# threshold the image and search for contours.
img = cv2.imread('tablelines.png')
res = img.copy()
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
_, thresh = cv2.threshold(gray,0,255,cv2.THRESH_BINARY_INV+cv2.THRESH_OTSU)
_, contours, hierarchy = cv2.findContours(thresh,cv2.RETR_TREE,cv2.CHAIN_APPROX_NONE)

# Iterate through contours and draw a slightly bigger white rectangle
# over the contours that are not big enough (the text) on the copy of the image.
for i in contours:
    cnt = cv2.contourArea(i)
    if cnt < 500:
        x,y,w,h = cv2.boundingRect(i)
        cv2.rectangle(res,(x-1,y-1),(x+w+1,y+h+1),(255,255,255),-1)

# Display the result. Note that the image is allready the template!
cv2.imshow('res', res)
cv2.waitKey(0)
cv2.destroyAllWindows()

# Optional count the rows and columns of the table
count = res.copy()
gray = cv2.cvtColor(count, cv2.COLOR_BGR2GRAY)
_, thresh = cv2.threshold(gray,0,255,cv2.THRESH_BINARY+cv2.THRESH_OTSU)
_, contours, hierarchy = cv2.findContours(thresh,cv2.RETR_TREE,cv2.CHAIN_APPROX_NONE)

check = []
for i in contours:
    cnt = cv2.contourArea(i)
    if 10000 > cnt > 10:
        cv2.drawContours(count, [i], 0, (255,255,0), 2)
        M = cv2.moments(i)
        cx = int(M['m10']/M['m00'])
        cy = int(M['m01']/M['m00'])
        check.append([cx, cy])

check.sort(key = lambda xy: xy[1])
columns = 1

for i in range(0, len(check)-1):
    if check[i+1][1] + 5 >= check[i][1] >= check[i+1][1] - 5:
        columns += 1
    else:
        break
print(columns)

check.sort(key = lambda tup: tup[0])
rows = 1
for i in range(0, len(check)-1):
    if check[i+1][0] + 5 >= check[i][0] >= check[i+1][0] - 5:
        rows += 1
    else:
        break
print('Columns: ',columns)
print('Roiws : ',rows)

cv2.imshow('res', count)
cv2.waitKey(0)
cv2.destroyAllWindows()



### LINES WITH HOUGHLINES()

# Convert the resulting image from previous step (no text) to gray colorspace.
res2 = img.copy()
gray = cv2.cvtColor(res, cv2.COLOR_BGR2GRAY)

# You can either use threshold or Canny edge for HoughLines().
_, thresh = cv2.threshold(gray,0,255,cv2.THRESH_BINARY_INV+cv2.THRESH_OTSU)
#edges = cv2.Canny(gray, 50, 150, apertureSize=3)

# Perform HoughLines tranform.  
lines = cv2.HoughLines(thresh,1,np.pi/180,200)
for line in lines:
    for rho,theta in line:
            a = np.cos(theta)
            b = np.sin(theta)
            x0 = a*rho
            y0 = b*rho
            x1 = int(x0 + 1000*(-b))
            y1 = int(y0 + 1000*(a))
            x2 = int(x0 - 1000*(-b))
            y2 = int(y0 - 1000*(a))

            cv2.line(res2,(x1,y1),(x2,y2),(0,0,255),2)

#Display the result.
cv2.imshow('res', res)
cv2.imshow('res2', res2)
cv2.waitKey(0)
cv2.destroyAllWindows()


### LINES WITH HOUGHLINESP()

# Convert the resulting image from first step (no text) to gray colorspace.
res3 = img.copy()
gray = cv2.cvtColor(res, cv2.COLOR_BGR2GRAY)

# Use Canny edge detection and dilate the edges for better result.
edges = cv2.Canny(gray, 50, 150, apertureSize=3)
kernel = np.ones((4,4),np.uint8)
dilation = cv2.dilate(edges,kernel,iterations = 1)

# Perform HoughLinesP tranform.  
minLineLength = 100
maxLineGap = 10
lines = cv2.HoughLinesP(dilation, 1, np.pi / 180, 50, minLineLength, maxLineGap)
for line in lines:
    for x1, y1, x2, y2 in line:
        cv2.line(res3, (x1, y1), (x2, y2), (0, 255, 0), 2)

#Display the result.
cv2.imwrite('h_res1.png', res)
cv2.imwrite('h_res2.png', res2)
cv2.imwrite('h_res3.png', res3)

cv2.imshow('res', res)
cv2.imshow('res2', res2)
cv2.imshow('res3', res3)
cv2.waitKey(0)
cv2.destroyAllWindows()

结果：

在此处输入图像描述

在此处输入图像描述

在此处输入图像描述

列数：7

行数：21

于 2018-10-15T16:32:40.073 回答
0

我在文档中找到了这个。

for x1,y1,x2,y2 in lines[0]:
    cv2.line(img,(x1,y1),(x2,y2),(0,255,0),2)

我不确定 的确切结构，lines但我看到的示例仅遍历第一个lines[0]. 也许您从for line in lines.

于 2018-10-15T12:22:28.610 回答