Question

我有以下两个链接：

https://pbs.twimg.com/media/Dv3pIsIUwAEdu--.jpg:large

https://pbs.twimg.com/media/Dv3lKfjV4AAkIpY.jpg:large

数据以表格格式显示，但是作为jpeg，我想捕获此信息并将其转换为df或小标题。

我尝试使用tesseract，但效果不佳，我的代码如下：

library(tesseract)
text <- ocr_data(input_1, engine = eng)
text <- tesseract::ocr_data("https://pbs.twimg.com/media/Dv3lKfjV4AAkIpY.jpg:large", engine = eng)

有什么想法吗？

Answer 1

尝试一些先决条件，例如转换为黑/白并删除网格。这应该可以帮助您开始：

library(magrittr)
library(magick)
#> Linking to ImageMagick 6.9.9.38
#> Enabled features: cairo, fontconfig, freetype, fftw, ghostscript, lcms, pango, rsvg, webp, x11
#> Disabled features:

# download file
url <- "https://pbs.twimg.com/media/Dv3pIsIUwAEdu--.jpg:large"
download.file(url, destfile = "table.jpg")

# convert to black and white
convert_bw <- 'convert table.jpg -fill white -fuzz 20% +opaque "#000000" table_bw.jpg'
system(convert_bw)

# remove grid
remove_grid <- "convert table_bw.jpg -negate -define morphology:compose=darken -morphology Thinning 'Rectangle:1x80+0+0<' -negate table_wo_grid.jpg"
system(remove_grid)

# read img and ocr
data <- image_read("table_wo_grid.jpg") %>%
  image_crop(geometry_area(0, 0, 80, 25)) %>%
  image_ocr() %>%
  stringi::stri_split(fixed = "\n")

head(data[[1]])
#> [1] "10/3/2013 112.32 -0.12 0.11 0.04 0.55 0.05 0.45 555 155 5.55 143,115 23,439 505"         
#> [2] "10/5/2013 112.94 -0.44 0.15 0.04 0.53 0.05 0.45 1,572 2,255 0.75 143,091 23,335 504"     
#> [3] "10/4/2013 115.53 -0.47 0.10 0.04 0.55 0.05 0.45 27,212 4,955 775,473 142,357 27,334 5 22"
#> [4] "10/5/2013 115.35 -0.57 0.00 0.04 0.51 0.05 0.29 25,522 5,312 4.05 131,320 25,340 513"    
#> [5] "10/2/2013 114.42 -0.51 0.01 0.04 0.44 0.05 0.19 470 994 0.47 121,250 25,901 74.53"       
#> [6] "9/23/2013 11495 -0.03 0.07 0.04 0.57 0.05 0.11 20,075 594 50 55 121,437 25,341 774773"

^{由reprex package（v0.2.1）于2019-01-02创建}

编辑无需系统调用即可进行转换

library(magrittr)
library(magick)
#> Linking to ImageMagick 6.9.9.38
#> Enabled features: cairo, fontconfig, freetype, fftw, ghostscript, lcms, pango, rsvg, webp, x11
#> Disabled features:

# download file
url <- "https://pbs.twimg.com/media/Dv3pIsIUwAEdu--.jpg:large"
download.file(url, destfile = "table.jpg")

# preprocessing
img <- image_read("table.jpg") %>% 
  image_transparent("white", fuzz=82) %>% 
  image_background("white") %>%
  image_negate() %>%
  image_morphology(method = "Thinning", kernel = "Rectangle:20x1+0+0^<") %>%
  image_negate() %>%
  image_crop(geometry_area(0, 0, 80, 25)) 

img

# read img and ocr
data <- img %>%
  image_ocr() 

# some wrangling
data %>%
  stringi::stri_split(fixed = "\n") %>%
  purrr::map(~ stringi::stri_split(str = ., fixed = "‘")) %>%
  .[[1]] %>%
  purrr::map_df(~ tibble::tibble(Date = .[1], Price = .[2], Change = .[3])) %>%
  dplyr::glimpse()
#> Observations: 61
#> Variables: 3
#> $ Date   <chr> "10/3/2013", "10/5/2013", "10/4/2013", "10/5/2013", "10...
#> $ Price  <chr> "11232", "11294", "11553", "11535", "114.42", "11495", ...
#> $ Change <chr> " -0.12", " -0.44", " -0.47", " -0.57", " -0.51", " -0....

^{由reprex package（v0.2.1）于2019-01-03创建}

将jpeg中的表提取到R中的数据框中

1 个答案: