使用 OpenCV 检测 .pdf 表单图像中的水平空白行

问题描述

我有 .pdf 文件已转换为该项目的 .jpg 图像.我的目标是识别您通常会在 .pdf 表单中找到的空白(例如 ____________),这些空白指示用户填写某种信息的空间.我一直在使用 cv2.Canny()cv2.HoughlinesP() 函数进行边缘检测.

I have .pdf files that have been converted to .jpg images for this project. My goal is to identify the blanks (e.g ____________) that you would generally find in a .pdf form that indicate a space for the user to sign of fill out some kind of information. I have been using edge detection with the cv2.Canny() and cv2.HoughlinesP() functions.

这工作得相当好,但有不少误报似乎不知从何而来.当我查看边缘"文件时,它会在其他单词周围显示一堆噪音.我不确定这种噪音是从哪里来的.

This works fairly well, but there are quite a few false positives that come about from seemingly nowhere. When I look at the 'edges' file it shows a bunch of noise around the other words. I'm uncertain where this noise comes from.

是否应该继续调整参数,还是有更好的方法来找到这些空白的位置?

Should I continue to tweak the parameters, or is there a better method to find the location of these blanks?


解决方案

假设您要在 .pdf 表单上查找水平线,这里有一个简单的方法:

Assuming that you're trying to find horizontal lines on a .pdf form, here's a simple approach:

  • 将图像转换为灰度和自适应阈值图像
  • 构造特殊内核以仅检测水平线
  • 执行形态转换
  • 查找轮廓并在图像上绘制

使用此示例图片

转换为灰度和自适应阈值得到二值图像

Convert to grayscale and adaptive threshold to obtain a binary image

gray = cv2.cvtColor(image,cv2.COLOR_BGR2GRAY)
thresh = cv2.threshold(gray, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)[1]

然后我们用 cv2.getStructuringElement() 创建一个内核,并进行形态变换以隔离水平线

Then we create a kernel with cv2.getStructuringElement() and perform morphological transformations to isolate horizontal lines

horizontal_kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (15,1))
detected_lines = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, horizontal_kernel, iterations=2)

从这里我们可以使用 cv2.HoughLinesP() 来检测线条,但是由于我们已经对图像进行了预处理并隔离了水平线,所以我们可以找到轮廓并绘制结果

From here we can use cv2.HoughLinesP() to detect lines but since we have already preprocessed the image and isolated the horizontal lines, we can just find contours and draw the result

cnts = cv2.findContours(detected_lines, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0] if len(cnts) == 2 else cnts[1]

for c in cnts:
    cv2.drawContours(image, [c], -1, (36,255,12), 3)

完整代码

import cv2

image = cv2.imread('2.png')
gray = cv2.cvtColor(image,cv2.COLOR_BGR2GRAY)
thresh = cv2.threshold(gray, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)[1]

horizontal_kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (15,1))
detected_lines = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, horizontal_kernel, iterations=2)

cnts = cv2.findContours(detected_lines, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0] if len(cnts) == 2 else cnts[1]

for c in cnts:
    cv2.drawContours(image, [c], -1, (36,255,12), 3)

cv2.imshow('thresh', thresh)
cv2.imshow('detected_lines', detected_lines)
cv2.imshow('image', image)
cv2.waitKey()

相关文章