Class: UdiseCaptchaReader::CharacterRecognizer

Inherits:
Object
  • Object
show all
Defined in:
lib/udise_captcha_reader/character_recognizer.rb

Constant Summary collapse

CHAR_SIZE =

Target size for normalized characters

32

Instance Method Summary collapse

Constructor Details

#initialize(options = {}) ⇒ CharacterRecognizer

Returns a new instance of CharacterRecognizer.



8
9
10
11
12
13
14
15
 
# File 'lib/udise_captcha_reader/character_recognizer.rb', line 8

def initialize(options = {})
  @options = {
    lang: "eng",
    processor: "mini_magick",
    psm: 10,
    oem: 1
  }.merge(options)
end

Instance Method Details

#recognize_character(image_file) ⇒ Object 



17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
 
# File 'lib/udise_captcha_reader/character_recognizer.rb', line 17

def recognize_character(image_file)
  image = MiniMagick::Image.open(image_file.path)
  
  # Normalize the character size
  width = image.width
  height = image.height
  scale = CHAR_SIZE / [width, height].max.to_f
  
  # Create a normalized version
  image.combine_options do |b|
    # Resize to standard size
    b.resize "#{(width * scale).round}x#{(height * scale).round}"
    
    # Center in a fixed-size canvas
    b.background "white"
    b.gravity "center"
    b.extent "#{CHAR_SIZE}x#{CHAR_SIZE}"
    
    # Convert to grayscale and adjust contrast
    b.colorspace "gray"
    b.brightness_contrast "10x20"  # Slight brightness boost and contrast
    
    # Clean threshold
    b.black_threshold "80%"
    b.white_threshold "20%"
  end
  
  # Save normalized image
  temp_output = Tempfile.new(['norm_char', '.png'])
  image.write(temp_output.path)
  
  # Try recognition with specific configurations
  results = []
  
  # Configuration attempts
  configs = [
    { psm: 10, whitelist: "0123456789abcdef" },
    { psm: 10, whitelist: "0123456789" },  # Numbers only
    { psm: 10, whitelist: "abcdef" }  # Hex letters only
  ]
  
  configs.each do |config|
    result = RTesseract.new(
      temp_output.path,
      lang: @options[:lang],
      processor: @options[:processor],
      psm: config[:psm],
      oem: 1,
      options: {
        "tessedit_char_whitelist" => config[:whitelist],
        "tessedit_ocr_engine_mode" => "1",
        "tessedit_pageseg_mode" => config[:psm].to_s,
        "tessdata_dir" => "/usr/local/share/tessdata",  # Ensure using latest tessdata
        "load_system_dawg" => "0",  # Disable dictionary
        "load_freq_dawg" => "0"     # Disable frequency-based corrections
      }
    ).to_s.strip.downcase
    
    results << result unless result.empty?
  end
  
  # Try with inverted image if no results
  if results.empty?
    image.negate
    image.write(temp_output.path)
    
    configs.each do |config|
      result = RTesseract.new(
        temp_output.path,
        lang: @options[:lang],
        processor: @options[:processor],
        psm: config[:psm],
        oem: 1,
        options: {
          "tessedit_char_whitelist" => config[:whitelist],
          "tessedit_ocr_engine_mode" => "1",
          "tessedit_pageseg_mode" => config[:psm].to_s,
          "tessdata_dir" => "/usr/local/share/tessdata",
          "load_system_dawg" => "0",
          "load_freq_dawg" => "0"
        }
      ).to_s.strip.downcase
      
      results << result unless result.empty?
    end
  end
  
  # Clean up
  temp_output.unlink
  
  # Process results
  char_counts = Hash.new(0)
  results.each do |result|
    next if result.empty?
    char = result[0].downcase
    # Map commonly confused characters
    char = '1' if char == 'i' || char == 'l' || char == '|'
    char = '0' if char == 'o'
    char_counts[char] += 1 if char.match?(/^[a-f0-9]/)
  end
  
  return '' if char_counts.empty?
  char_counts.max_by { |_, count| count }[0]
end