Web Scraping - 带有日期选择器的表格作为图像

Question

我是 Web Scraping 的新手。我想将表中的值提取到 Pandas 中。表格是一张图片，即无法通过选择右键单击并复制和粘贴数据。我该怎么做？我无法“识别”表格图像，控制台输出的简单 ctrl+f 似乎没有返回 table 中的任何值。

from bs4 import BeautifulSoup as bs
import pandas as pd
pd.set_option('display.max_colwidth', 500)
import time
import requests
import pytesseract as pytes
import random
myURL =  "https://www.netztransparenz.de/EEG/Marktpraemie/EPEX-SPOT-Stundenkontrakte"
page = requests.get(myURL)

print (page)

soup = bs(page.content,"html.parser")

print(soup) <--- ctrl + F the result to locate values seen in the table

Answer 1

我自己也去了。几乎是我第一次使用pytesseract，所以把它当作它的价值：o)...

查看页面来源，可以从url获取表格图片：

https://www.netztransparenz.de/DesktopModules/LotesNetztransparenz/ImageCharts/EpexChartImageHandler.ashx?date=2021/08/23&type=1

网址的后半部分（不完全在末尾）有日期：date=2021/08/23。如果这被改变，数据的时间段就会改变，你会得到一个新的图像。因此，以下代码似乎纯粹是作为尝试访问三个图像的检查：

import urllib.request
from time import sleep

main_url = 'https://www.netztransparenz.de/DesktopModules/LotesNetztransparenz/ImageCharts/EpexChartImageHandler.ashx?date='

tail_url = '&type=1'

dates = ['2021/08/21', '2021/08/22', '2021/08/23']

for date in dates:
   r = urllib.request.urlopen(main_url+date+tail_url)
   print(r.getcode())
   sleep(1)
   
>> 200 #'200' is 'OK'
>> 200
>> 200

无论如何，对于一个 url，以下代码会接近：

# https://pypi.org/project/opencv-python/
import cv2
import urllib.request  #I think (?) cv2 wants/prefers urllib instead of requests
import numpy as np

# sudo apt-get update
# sudo apt install tesseract-ocr
# sudo apt install libtesseract-dev
# ...or find for Mac/Windows.  Then...
# https://pypi.org/project/pytesseract/
import pytesseract
import pandas as pd
from io import StringIO

# first part of the url
main_url = 'https://www.netztransparenz.de/DesktopModules/LotesNetztransparenz/ImageCharts/EpexChartImageHandler.ashx?date='

# end part of the url
tail_url = '&type=1'

# date part - the code could be expanded to loop different dates
# and get different frames and concat them together.  For demo let's just get just one
date = '2021/08/23'

#construction a valid url.  If looping and the 'date' above was a list of dates this method would make more sense
url = main_url + date + tail_url

# this probably wants a try except around it and proceed only on an OK/200 response
url_response = urllib.request.urlopen(url)

# download the image
img = cv2.imdecode(np.array(bytearray(url_response.read()), dtype=np.uint8), -1)

# get the image into text
text = pytesseract.image_to_string(img)

# create a rough frame and clean up
df = pd.read_csv(StringIO(text), sep=';', header=0)
df = df[:-2]
# df['a'].str.rsplit(n=7, expand=True)
df['Stunden'] = df['a'].str.extract(r'((?:\d{2}:?\d{2}) - (?:\d{2}:?\d{2}))\s')
df = df.set_index(df['Stunden'])
df['a'] = df['a'].str.replace(r'((?:\d{2}:?\d{2}) - (?:\d{2}:?\d{2}))\s', '', regex=True)
df = df['a'].str.rsplit(n=7, expand=True)
df = df.reset_index()

# create a dict for column renaming
datesplit = [int(x) for x in date.split('/')]
weekdates = {}
count = 0
for i in range(7, -1, -1):
    d = datesplit[2] - i
    weekdates[count] = (str(datesplit[0]) + '-' + str(datesplit[1]) + '-' + str(d))
    count += 1

# rename columns
df = df.rename(columns=weekdates)

# output frame
display(df)

看起来还不错。 然而，如果你仔细观察，它并不完美......

如果帧在图像处理中只有一些小故障，您可以使用以下行手动纠正它们：

df.loc[12, '2021-8-20'] = '9,911'
df.loc[17, 'Stunden'] = '17:00 - 18:00'

但是，如果有更多问题（并且可能存在），最好完全替换点和逗号和/或尝试更广泛的清理策略。或者希望改进图像的处理（如果您遇到困难，可以发布另一个问题，以及对模块有更多了解的人的回答 pytesseract 和/或 cv2）。

顺便说一句：您可以看出问题出在图像处理中，因为检查 text 变量会显示在将其放入数据帧之前最初提取的内容...

Answer 2

我使用 pytesseract 解决了这个问题。可以在这里找到有用的建议。 Use pytesseract OCR to recognize text from an image