H5 图像识别

客官°小女子只卖身不卖艺 2022-06-04 03:09 281阅读 0赞

#### **识别对比** ####

--------------------

##### **1、百度识别** #####

发现百度的图片搜索识别率不是特别，下面为测试图片跟测试后的结果：

测试图片：![图片1][1]

下面为测试后的结果：

![图片2][2]

![图片3][3]

##### **2、采用 [tesseract.js][] 后结果** #####

![对比图][SouthEast]

#### **H5 图像识别 （采用[Tesseract.js][tesseract.js] 进行识别）** ####

--------------------

简单的文案之类的，识别的还算可以，但是稍微复杂点的，准确率就不是那么好了，在学习中。。。

#### **安装** ####

或者

> npm install tesseract.js –save

PS：如果使用 npm 安装异常，可以使用 cnpm 进行安装使用

#### **使用** ####

**demo 1：then使用**

var Tesseract = require('tesseract.js')
    
    Tesseract.recognize(myImage).then(function(result){ 
        console.log(result)
    })

**demo 2：lang切换**

Tesseract.recognize(myImage, {
        lang: 'spa',
        tessedit_char_blacklist: 'e'
    }).then(function(result){ 
        console.log(result)
    })

**demo 3：（then、progress、catch、then、finally）**

Tesseract.recognize(src, {
            lang:"chi_sim",
        })
        .progress(function(message) { 
            console.log(message)
        })
        .catch(function(err) { 
            console.error(err)
        })
        .then(function(result) { 
            console.log(result.text)
        })
        .finally(function(resultOrError) { 
            console.log(resultOrError)
        })

#### **参数介绍：** ####

**1、image是任何 参数介绍：**

image是任何 [ImageLike][] 对象，取决于它是从浏览器还是通过NodeJS运行。

第一个参数，可以是 img 路劲地址，可以是图片base64位的二进制码、也可以是Image对象 等。

附上实现的代码：

<!DOCTYPE html>
    <html>
    
        <head>
            <meta charset="UTF-8">
            <meta name="viewport" content="width=device-width,initial-scale=1,shrink-to-fit=no,user-scalable=no,minimum-scale=1,maximum-scale=1">
            <title>图片识别</title>
            <style>body{ margin:0 auto;width:500px;font-size:12px;font-family:"arial, helvetica, sans-serif"}fieldset{ margin-bottom:10%;border:1px solid #ddd;border-radius:5px}img,select,button{ cursor:pointer}img{ background:#ddd}h2{ font-weight:500;font-size:16px}fieldset legend{ margin-left:33%}</style>
        </head>
    
        <body>
    
            <fieldset>
                <legend>
                    <h2> 图片识别前 </h2>
                </legend>
                选择文件：<img src="img/1.png" title="图片" />
                <p>
                    选择语言：
                    <select id="langsel" onchange="recognizeFile()">
                        <option value='afr'> 南非荷兰语（Afrikaans） </option>
                        <option value='ara'> 阿拉伯 （Arabic）</option>
                        <option value='aze'> 阿塞拜疆 （Azerbaijani） </option>
                        <option value='bel'> 白俄罗斯（美式：Belarusian） </option>
                        <option value='ben'> 孟加拉（Bengali） </option>
                        <option value='bul'> 保加利亚语（Bulgarian） </option>
                        <option value='cat'> 西班牙 （Catalan）</option>
                        <option value='ces'> 捷克（Czech） </option>
                        <option value='chi_sim' selected> 中文（Chinese） </option>
                        <option value='chi_tra'> 繁体中文（Traditional Chinese） </option>
                        <option value='chr'> Cherokee </option>
                        <option value='dan'> 丹麦语（Danish） </option>
                        <option value='deu'> 德语（German） </option>
                        <option value='ell'> 希腊语（Greek） </option>
                        <option value='eng'> 英语（English） </option>
                        <option value='enm'> 英文（老）（English (Old)） </option>
                        <option value='meme'> Internet Meme </option>
                        <option value='epo'> Esperanto </option>
                        <option value='epo_alt'> （Esperanto alternative）</option>
                        <option value='equ'> Math </option>
                        <option value='est'> Estonian </option>
                        <option value='eus'> 爱沙尼亚语（Basque） </option>
                        <option value='fin'> （Finnish） </option>
                        <option value='fra'> 芬兰（French） </option>
                        <option value='frk'> Frankish </option>
                        <option value='frm'> 法语（老）（French (Old)） </option>
                        <option value='glg'> 加利西亚（Galician） </option>
                        <option value='grc'> Ancient Greek </option>
                        <option value='heb'> Hebrew </option>
                        <option value='hin'> Hindi </option>
                        <option value='hrv'> Croatian </option>
                        <option value='hun'> Hungarian </option>
                        <option value='ind'> Indonesian </option>
                        <option value='isl'> Icelandic </option>
                        <option value='ita'> 意大利（Italian） </option>
                        <option value='ita_old'> 意大利语（老）（Italian (Old)） </option>
                        <option value='jpn'> 日本（Japanese） </option>
                        <option value='kan'> Kannada </option>
                        <option value='kor'> 朝鲜的（Korean） </option>
                        <option value='lav'> Latvian </option>
                        <option value='lit'> Lithuanian </option>
                        <option value='mal'> Malayalam </option>
                        <option value='mkd'> Macedonian </option>
                        <option value='mlt'> Maltese </option>
                        <option value='msa'> Malay </option>
                        <option value='nld'> Dutch </option>
                        <option value='nor'> Norwegian </option>
                        <option value='pol'> Polish </option>
                        <option value='por'> 葡萄牙语（Portuguese） </option>
                        <option value='ron'> Romanian </option>
                        <option value='rus'> 俄语（Russian） </option>
                        <option value='slk'> Slovakian </option>
                        <option value='slv'> Slovenian </option>
                        <option value='spa'> 西班牙语（Spanish） </option>
                        <option value='spa_old'>老西班牙语 （Old Spanish） </option>
                        <option value='sqi'> Albanian </option>
                        <option value='srp'> 塞尔维亚语（拉丁语）（Serbian (Latin)） </option>
                        <option value='swa'> Swahili </option>
                        <option value='swe'> Swedish </option>
                        <option value='tam'> Tamil </option>
                        <option value='tel'> Telugu </option>
                        <option value='tgl'> Tagalog </option>
                        <option value='tha'> 泰国（Thai） </option>
                        <option value='tur'> 土耳其（Turkish） </option>
                        <option value='ukr'> 乌克兰（乌克兰） </option>
                        <option value='vie'> 越南（Vietnamese） </option>
                    </select>
                </p>
                <p align="center">
                    <button onclick="btn()">执行</button>
                </p>
            </fieldset>
            <fieldset>
                <legend>
                    <h2> 输出结果 </h2>
                </legend>
                <div id="result"></div>
            </fieldset>
        </body>
        <script src='img/tesseract.js'></script>
        <script> var src = document.querySelector("img").src, selectOption = "", result = document.querySelector("#result"); function recognizeFile() { var select = document.querySelector("#langsel") selectOption = select.options[select.selectedIndex].value; } function btn() { Tesseract.recognize(src, { lang: selectOption ? selectOption : "chi_sim", }).progress(function(message) { console.log(message) }) .catch(function(err) { result.innerHTML = err; console.error(err) }) .then(function(result) { console.log(result.text) result.value = result.text; }) .finally(function(resultOrError) { result.innerHTML = resultOrError.value; console.log(resultOrError) }) } </script>
    
    </html>

**2、语言支持介绍：**

<table> 
 <thead> 
  <tr> 
   <th><code>lang</code></th> 
   <th>Language</th> 
  </tr> 
 </thead> 
 <tbody> 
  <tr> 
   <td>‘afr’</td> 
   <td>Afrikaans</td> 
  </tr> 
  <tr> 
   <td>‘ara’</td> 
   <td>Arabic</td> 
  </tr> 
  <tr> 
   <td>‘aze’</td> 
   <td>Azerbaijani</td> 
  </tr> 
  <tr> 
   <td>‘bel’</td> 
   <td>Belarusian</td> 
  </tr> 
  <tr> 
   <td>‘ben’</td> 
   <td>Bengali</td> 
  </tr> 
  <tr> 
   <td>‘bul’</td> 
   <td>Bulgarian</td> 
  </tr> 
  <tr> 
   <td>‘cat’</td> 
   <td>Catalan</td> 
  </tr> 
  <tr> 
   <td>‘ces’</td> 
   <td>Czech</td> 
  </tr> 
  <tr> 
   <td>‘chi_sim’</td> 
   <td>Chinese</td> 
  </tr> 
  <tr> 
   <td>‘chi_tra’</td> 
   <td>Traditional Chinese</td> 
  </tr> 
  <tr> 
   <td>‘chr’</td> 
   <td>Cherokee</td> 
  </tr> 
  <tr> 
   <td>‘dan’</td> 
   <td>Danish</td> 
  </tr> 
  <tr> 
   <td>‘deu’</td> 
   <td>German</td> 
  </tr> 
  <tr> 
   <td>‘ell’</td> 
   <td>Greek</td> 
  </tr> 
  <tr> 
   <td>‘eng’</td> 
   <td>English</td> 
  </tr> 
  <tr> 
   <td>‘enm’</td> 
   <td>English (Old)</td> 
  </tr> 
  <tr> 
   <td>‘epo’</td> 
   <td>Esperanto</td> 
  </tr> 
  <tr> 
   <td>‘epo_alt’</td> 
   <td>Esperanto alternative</td> 
  </tr> 
  <tr> 
   <td>‘equ’</td> 
   <td>Math</td> 
  </tr> 
  <tr> 
   <td>‘est’</td> 
   <td>Estonian</td> 
  </tr> 
  <tr> 
   <td>‘eus’</td> 
   <td>Basque</td> 
  </tr> 
  <tr> 
   <td>‘fin’</td> 
   <td>Finnish</td> 
  </tr> 
  <tr> 
   <td>‘fra’</td> 
   <td>French</td> 
  </tr> 
  <tr> 
   <td>‘frk’</td> 
   <td>Frankish</td> 
  </tr> 
  <tr> 
   <td>‘frm’</td> 
   <td>French (Old)</td> 
  </tr> 
  <tr> 
   <td>‘glg’</td> 
   <td>Galician</td> 
  </tr> 
  <tr> 
   <td>‘grc’</td> 
   <td>Ancient Greek</td> 
  </tr> 
  <tr> 
   <td>‘heb’</td> 
   <td>Hebrew</td> 
  </tr> 
  <tr> 
   <td>‘hin’</td> 
   <td>Hindi</td> 
  </tr> 
  <tr> 
   <td>‘hrv’</td> 
   <td>Croatian</td> 
  </tr> 
  <tr> 
   <td>‘hun’</td> 
   <td>Hungarian</td> 
  </tr> 
  <tr> 
   <td>‘ind’</td> 
   <td>Indonesian</td> 
  </tr> 
  <tr> 
   <td>‘isl’</td> 
   <td>Icelandic</td> 
  </tr> 
  <tr> 
   <td>‘ita’</td> 
   <td>Italian</td> 
  </tr> 
  <tr> 
   <td>‘ita_old’</td> 
   <td>Italian (Old)</td> 
  </tr> 
  <tr> 
   <td>‘jpn’</td> 
   <td>Japanese</td> 
  </tr> 
  <tr> 
   <td>‘kan’</td> 
   <td>Kannada</td> 
  </tr> 
  <tr> 
   <td>‘kor’</td> 
   <td>Korean</td> 
  </tr> 
  <tr> 
   <td>‘lav’</td> 
   <td>Latvian</td> 
  </tr> 
  <tr> 
   <td>‘lit’</td> 
   <td>Lithuanian</td> 
  </tr> 
  <tr> 
   <td>‘mal’</td> 
   <td>Malayalam</td> 
  </tr> 
  <tr> 
   <td>‘mkd’</td> 
   <td>Macedonian</td> 
  </tr> 
  <tr> 
   <td>‘mlt’</td> 
   <td>Maltese</td> 
  </tr> 
  <tr> 
   <td>‘msa’</td> 
   <td>Malay</td> 
  </tr> 
  <tr> 
   <td>‘nld’</td> 
   <td>Dutch</td> 
  </tr> 
  <tr> 
   <td>‘nor’</td> 
   <td>Norwegian</td> 
  </tr> 
  <tr> 
   <td>‘pol’</td> 
   <td>Polish</td> 
  </tr> 
  <tr> 
   <td>‘por’</td> 
   <td>Portuguese</td> 
  </tr> 
  <tr> 
   <td>‘ron’</td> 
   <td>Romanian</td> 
  </tr> 
  <tr> 
   <td>‘rus’</td> 
   <td>Russian</td> 
  </tr> 
  <tr> 
   <td>‘slk’</td> 
   <td>Slovakian</td> 
  </tr> 
  <tr> 
   <td>‘slv’</td> 
   <td>Slovenian</td> 
  </tr> 
  <tr> 
   <td>‘spa’</td> 
   <td>Spanish</td> 
  </tr> 
  <tr> 
   <td>‘spa_old’</td> 
   <td>Old Spanish</td> 
  </tr> 
  <tr> 
   <td>‘sqi’</td> 
   <td>Albanian</td> 
  </tr> 
  <tr> 
   <td>‘srp’</td> 
   <td>Serbian (Latin)</td> 
  </tr> 
  <tr> 
   <td>‘swa’</td> 
   <td>Swahili</td> 
  </tr> 
  <tr> 
   <td>‘swe’</td> 
   <td>Swedish</td> 
  </tr> 
  <tr> 
   <td>‘tam’</td> 
   <td>Tamil</td> 
  </tr> 
  <tr> 
   <td>‘tel’</td> 
   <td>Telugu</td> 
  </tr> 
  <tr> 
   <td>‘tgl’</td> 
   <td>Tagalog</td> 
  </tr> 
  <tr> 
   <td>‘tha’</td> 
   <td>Thai</td> 
  </tr> 
  <tr> 
   <td>‘tur’</td> 
   <td>Turkish</td> 
  </tr> 
  <tr> 
   <td>‘ukr’</td> 
   <td>Ukrainian</td> 
  </tr> 
  <tr> 
   <td>‘vie’</td> 
   <td>Vietnamese</td> 
  </tr> 
 </tbody> 
</table>

**Tesseract参数支持介绍：**

<table> 
 <thead> 
  <tr> 
   <th>Parameter</th> 
   <th>Default Value</th> 
   <th>Description</th> 
  </tr> 
 </thead> 
 <tbody> 
  <tr> 
   <td>ambigs_debug_level</td> 
   <td>0</td> 
   <td>Debug level for unichar ambiguities</td> 
  </tr> 
  <tr> 
   <td>applybox_debug</td> 
   <td>1</td> 
   <td>Debug level</td> 
  </tr> 
  <tr> 
   <td>applybox_exposure_pattern</td> 
   <td>.exp</td> 
   <td>Exposure value follows this pattern in the image filename. The name of the image files are expected to be in the form [lang].[fontname].exp[num].tif</td> 
  </tr> 
  <tr> 
   <td>applybox_learn_chars_and_char_frags_mode</td> 
   <td>0</td> 
   <td>Learn both character fragments (as is done in the special low exposure mode) as well as unfragmented characters.</td> 
  </tr> 
  <tr> 
   <td>applybox_learn_ngrams_mode</td> 
   <td>0</td> 
   <td>Each bounding box is assumed to contain ngrams. Only learn the ngrams whose outlines overlap horizontally.</td> 
  </tr> 
  <tr> 
   <td>applybox_page</td> 
   <td>0</td> 
   <td>Page number to apply boxes from</td> 
  </tr> 
  <tr> 
   <td>assume_fixed_pitch_char_segment</td> 
   <td>0</td> 
   <td>include fixed-pitch heuristics in char segmentation</td> 
  </tr> 
  <tr> 
   <td>bestrate_pruning_factor</td> 
   <td>2</td> 
   <td>Multiplying factor of current best rate to prune other hypotheses</td> 
  </tr> 
  <tr> 
   <td>bidi_debug</td> 
   <td>0</td> 
   <td>Debug level for BiDi</td> 
  </tr> 
  <tr> 
   <td>bland_unrej</td> 
   <td>0</td> 
   <td>unrej potential with no chekcs</td> 
  </tr> 
  <tr> 
   <td>certainty_scale</td> 
   <td>20</td> 
   <td>Certainty scaling factor</td> 
  </tr> 
  <tr> 
   <td>certainty_scale</td> 
   <td>20</td> 
   <td>Certainty scaling factor</td> 
  </tr> 
  <tr> 
   <td>chop_center_knob</td> 
   <td>0.15</td> 
   <td>Split center adjustment</td> 
  </tr> 
  <tr> 
   <td>chop_centered_maxwidth</td> 
   <td>90</td> 
   <td>Width of (smaller) chopped blobs above which we don’t care that a chop is not near the center.</td> 
  </tr> 
  <tr> 
   <td>chop_debug</td> 
   <td>0</td> 
   <td>Chop debug</td> 
  </tr> 
  <tr> 
   <td>chop_enable</td> 
   <td>1</td> 
   <td>Chop enable</td> 
  </tr> 
  <tr> 
   <td>chop_good_split</td> 
   <td>50</td> 
   <td>Good split limit</td> 
  </tr> 
  <tr> 
   <td>chop_inside_angle</td> 
   <td>-50</td> 
   <td>Min Inside Angle Bend</td> 
  </tr> 
  <tr> 
   <td>chop_min_outline_area</td> 
   <td>2000</td> 
   <td>Min Outline Area</td> 
  </tr> 
  <tr> 
   <td>chop_min_outline_points</td> 
   <td>6</td> 
   <td>Min Number of Points on Outline</td> 
  </tr> 
  <tr> 
   <td>chop_new_seam_pile</td> 
   <td>1</td> 
   <td>Use new seam_pile</td> 
  </tr> 
  <tr> 
   <td>chop_ok_split</td> 
   <td>100</td> 
   <td>OK split limit</td> 
  </tr> 
  <tr> 
   <td>chop_overlap_knob</td> 
   <td>0.9</td> 
   <td>Split overlap adjustment</td> 
  </tr> 
  <tr> 
   <td>chop_same_distance</td> 
   <td>2</td> 
   <td>Same distance</td> 
  </tr> 
  <tr> 
   <td>chop_seam_pile_size</td> 
   <td>150</td> 
   <td>Max number of seams in seam_pile</td> 
  </tr> 
  <tr> 
   <td>chop_sharpness_knob</td> 
   <td>0.06</td> 
   <td>Split sharpness adjustment</td> 
  </tr> 
  <tr> 
   <td>chop_split_dist_knob</td> 
   <td>0.5</td> 
   <td>Split length adjustment</td> 
  </tr> 
  <tr> 
   <td>chop_split_length</td> 
   <td>10000</td> 
   <td>Split Length</td> 
  </tr> 
  <tr> 
   <td>chop_vertical_creep</td> 
   <td>0</td> 
   <td>Vertical creep</td> 
  </tr> 
  <tr> 
   <td>chop_width_change_knob</td> 
   <td>5</td> 
   <td>Width change adjustment</td> 
  </tr> 
  <tr> 
   <td>chop_x_y_weight</td> 
   <td>3</td> 
   <td>X / Y length weight</td> 
  </tr> 
  <tr> 
   <td>chs_leading_punct</td> 
   <td>(‘`”</td> 
   <td>Leading punctuation</td> 
  </tr> 
  <tr> 
   <td>chs_trailing_punct1</td> 
   <td>).,;:?!</td> 
   <td>1st Trailing punctuation</td> 
  </tr> 
  <tr> 
   <td>chs_trailing_punct2</td> 
   <td>)’`”</td> 
   <td>2nd Trailing punctuation</td> 
  </tr> 
  <tr> 
   <td>classify_adapt_feature_threshold</td> 
   <td>230</td> 
   <td>Threshold for good features during adaptive 0-255</td> 
  </tr> 
  <tr> 
   <td>classify_adapt_proto_threshold</td> 
   <td>230</td> 
   <td>Threshold for good protos during adaptive 0-255</td> 
  </tr> 
  <tr> 
   <td>classify_adapted_pruning_factor</td> 
   <td>2.5</td> 
   <td>Prune poor adapted results this much worse than best result</td> 
  </tr> 
  <tr> 
   <td>classify_adapted_pruning_threshold</td> 
   <td>-1</td> 
   <td>Threshold at which classify_adapted_pruning_factor starts</td> 
  </tr> 
  <tr> 
   <td>classify_bln_numeric_mode</td> 
   <td>0</td> 
   <td>Assume the input is numbers [0-9].</td> 
  </tr> 
  <tr> 
   <td>classify_char_norm_range</td> 
   <td>0.2</td> 
   <td>Character Normalization Range …</td> 
  </tr> 
  <tr> 
   <td>classify_character_fragments_garbage_certainty_threshold</td> 
   <td>-3</td> 
   <td>Exclude fragments that do not look like whole characters from training and adaption</td> 
  </tr> 
  <tr> 
   <td>classify_class_pruner_multiplier</td> 
   <td>15</td> 
   <td>Class Pruner Multiplier 0-255:</td> 
  </tr> 
  <tr> 
   <td>classify_class_pruner_threshold</td> 
   <td>229</td> 
   <td>Class Pruner Threshold 0-255</td> 
  </tr> 
  <tr> 
   <td>classify_cp_angle_pad_loose</td> 
   <td>45</td> 
   <td>Class Pruner Angle Pad Loose</td> 
  </tr> 
  <tr> 
   <td>classify_cp_angle_pad_medium</td> 
   <td>20</td> 
   <td>Class Pruner Angle Pad Medium</td> 
  </tr> 
  <tr> 
   <td>classify_cp_angle_pad_tight</td> 
   <td>10</td> 
   <td>CLass Pruner Angle Pad Tight</td> 
  </tr> 
  <tr> 
   <td>classify_cp_cutoff_strength</td> 
   <td>7</td> 
   <td>Class Pruner CutoffStrength:</td> 
  </tr> 
  <tr> 
   <td>classify_cp_end_pad_loose</td> 
   <td>0.5</td> 
   <td>Class Pruner End Pad Loose</td> 
  </tr> 
  <tr> 
   <td>classify_cp_end_pad_medium</td> 
   <td>0.5</td> 
   <td>Class Pruner End Pad Medium</td> 
  </tr> 
  <tr> 
   <td>classify_cp_end_pad_tight</td> 
   <td>0.5</td> 
   <td>Class Pruner End Pad Tight</td> 
  </tr> 
  <tr> 
   <td>classify_cp_side_pad_loose</td> 
   <td>2.5</td> 
   <td>Class Pruner Side Pad Loose</td> 
  </tr> 
  <tr> 
   <td>classify_cp_side_pad_medium</td> 
   <td>1.2</td> 
   <td>Class Pruner Side Pad Medium</td> 
  </tr> 
  <tr> 
   <td>classify_cp_side_pad_tight</td> 
   <td>0.6</td> 
   <td>Class Pruner Side Pad Tight</td> 
  </tr> 
  <tr> 
   <td>classify_debug_character_fragments</td> 
   <td>0</td> 
   <td>Bring up graphical debugging windows for fragments training</td> 
  </tr> 
  <tr> 
   <td>classify_debug_level</td> 
   <td>0</td> 
   <td>Classify debug level</td> 
  </tr> 
  <tr> 
   <td>classify_enable_adaptive_debugger</td> 
   <td>0</td> 
   <td>Enable match debugger</td> 
  </tr> 
  <tr> 
   <td>classify_enable_adaptive_matcher</td> 
   <td>1</td> 
   <td>Enable adaptive classifier</td> 
  </tr> 
  <tr> 
   <td>classify_enable_learning</td> 
   <td>1</td> 
   <td>Enable adaptive classifier</td> 
  </tr> 
  <tr> 
   <td>classify_font_name</td> 
   <td>UnknownFont</td> 
   <td>Default font name to be used in training</td> 
  </tr> 
  <tr> 
   <td>classify_integer_matcher_multiplier</td> 
   <td>10</td> 
   <td>Integer Matcher Multiplier 0-255:</td> 
  </tr> 
  <tr> 
   <td>classify_learn_debug_str</td> 
   <td></td> 
   <td>Class str to debug learning</td> 
  </tr> 
  <tr> 
   <td>classify_learning_debug_level</td> 
   <td>0</td> 
   <td>Learning Debug Level:</td> 
  </tr> 
  <tr> 
   <td>classify_max_certainty_margin</td> 
   <td>5.5</td> 
   <td>Veto difference between classifier certainties</td> 
  </tr> 
  <tr> 
   <td>classify_max_norm_scale_x</td> 
   <td>0.325</td> 
   <td>Max char x-norm scale …</td> 
  </tr> 
  <tr> 
   <td>classify_max_norm_scale_y</td> 
   <td>0.325</td> 
   <td>Max char y-norm scale …</td> 
  </tr> 
  <tr> 
   <td>classify_max_rating_ratio</td> 
   <td>1.5</td> 
   <td>Veto ratio between classifier ratings</td> 
  </tr> 
  <tr> 
   <td>classify_max_slope</td> 
   <td>2.41421</td> 
   <td>Slope above which lines are called vertical</td> 
  </tr> 
  <tr> 
   <td>classify_min_norm_scale_x</td> 
   <td>0</td> 
   <td>Min char x-norm scale …</td> 
  </tr> 
  <tr> 
   <td>classify_min_norm_scale_y</td> 
   <td>0</td> 
   <td>Min char y-norm scale …</td> 
  </tr> 
  <tr> 
   <td>classify_min_slope</td> 
   <td>0.414214</td> 
   <td>Slope below which lines are called horizontal</td> 
  </tr> 
  <tr> 
   <td>classify_misfit_junk_penalty</td> 
   <td>0</td> 
   <td>Penalty to apply when a non-alnum is vertically out of its expected textline position</td> 
  </tr> 
  <tr> 
   <td>classify_nonlinear_norm</td> 
   <td>0</td> 
   <td>Non-linear stroke-density normalization</td> 
  </tr> 
  <tr> 
   <td>classify_norm_adj_curl</td> 
   <td>2</td> 
   <td>Norm adjust curl …</td> 
  </tr> 
  <tr> 
   <td>classify_norm_adj_midpoint</td> 
   <td>32</td> 
   <td>Norm adjust midpoint …</td> 
  </tr> 
  <tr> 
   <td>classify_norm_method</td> 
   <td>1</td> 
   <td>Normalization Method …</td> 
  </tr> 
  <tr> 
   <td>classify_num_cp_levels</td> 
   <td>3</td> 
   <td>Number of Class Pruner Levels</td> 
  </tr> 
  <tr> 
   <td>classify_pico_feature_length</td> 
   <td>0.05</td> 
   <td>Pico Feature Length</td> 
  </tr> 
  <tr> 
   <td>classify_pp_angle_pad</td> 
   <td>45</td> 
   <td>Proto Pruner Angle Pad</td> 
  </tr> 
  <tr> 
   <td>classify_pp_end_pad</td> 
   <td>0.5</td> 
   <td>Proto Prune End Pad</td> 
  </tr> 
  <tr> 
   <td>classify_pp_side_pad</td> 
   <td>2.5</td> 
   <td>Proto Pruner Side Pad</td> 
  </tr> 
  <tr> 
   <td>classify_save_adapted_templates</td> 
   <td>0</td> 
   <td>Save adapted templates to a file</td> 
  </tr> 
  <tr> 
   <td>classify_training_file</td> 
   <td>MicroFeatures</td> 
   <td>Training file</td> 
  </tr> 
  <tr> 
   <td>classify_use_pre_adapted_templates</td> 
   <td>0</td> 
   <td>Use pre-adapted classifier templates</td> 
  </tr> 
  <tr> 
   <td>conflict_set_I_l_1</td> 
   <td>Il1[]</td> 
   <td>Il1 conflict set</td> 
  </tr> 
  <tr> 
   <td>crunch_accept_ok</td> 
   <td>1</td> 
   <td>Use acceptability in okstring</td> 
  </tr> 
  <tr> 
   <td>crunch_debug</td> 
   <td>0</td> 
   <td>As it says</td> 
  </tr> 
  <tr> 
   <td>crunch_del_cert</td> 
   <td>-10</td> 
   <td>POTENTIAL crunch cert lt this</td> 
  </tr> 
  <tr> 
   <td>crunch_del_high_word</td> 
   <td>1.5</td> 
   <td>Del if word gt xht x this above bl</td> 
  </tr> 
  <tr> 
   <td>crunch_del_low_word</td> 
   <td>0.5</td> 
   <td>Del if word gt xht x this below bl</td> 
  </tr> 
  <tr> 
   <td>crunch_del_max_ht</td> 
   <td>3</td> 
   <td>Del if word ht gt xht x this</td> 
  </tr> 
  <tr> 
   <td>crunch_del_min_ht</td> 
   <td>0.7</td> 
   <td>Del if word ht lt xht x this</td> 
  </tr> 
  <tr> 
   <td>crunch_del_min_width</td> 
   <td>3</td> 
   <td>Del if word width lt xht x this</td> 
  </tr> 
  <tr> 
   <td>crunch_del_rating</td> 
   <td>60</td> 
   <td>POTENTIAL crunch rating lt this</td> 
  </tr> 
  <tr> 
   <td>crunch_early_convert_bad_unlv_chs</td> 
   <td>0</td> 
   <td>Take out ~^ early?</td> 
  </tr> 
  <tr> 
   <td>crunch_early_merge_tess_fails</td> 
   <td>1</td> 
   <td>Before word crunch?</td> 
  </tr> 
  <tr> 
   <td>crunch_include_numerals</td> 
   <td>0</td> 
   <td>Fiddle alpha figures</td> 
  </tr> 
  <tr> 
   <td>crunch_leave_accept_strings</td> 
   <td>0</td> 
   <td>Dont pot crunch sensible strings</td> 
  </tr> 
  <tr> 
   <td>crunch_leave_lc_strings</td> 
   <td>4</td> 
   <td>Dont crunch words with long lower case strings</td> 
  </tr> 
  <tr> 
   <td>crunch_leave_ok_strings</td> 
   <td>1</td> 
   <td>Dont touch sensible strings</td> 
  </tr> 
  <tr> 
   <td>crunch_leave_uc_strings</td> 
   <td>4</td> 
   <td>Dont crunch words with long lower case strings</td> 
  </tr> 
  <tr> 
   <td>crunch_long_repetitions</td> 
   <td>3</td> 
   <td>Crunch words with long repetitions</td> 
  </tr> 
  <tr> 
   <td>crunch_poor_garbage_cert</td> 
   <td>-9</td> 
   <td>crunch garbage cert lt this</td> 
  </tr> 
  <tr> 
   <td>crunch_poor_garbage_rate</td> 
   <td>60</td> 
   <td>crunch garbage rating lt this</td> 
  </tr> 
  <tr> 
   <td>crunch_pot_garbage</td> 
   <td>1</td> 
   <td>POTENTIAL crunch garbage</td> 
  </tr> 
  <tr> 
   <td>crunch_pot_indicators</td> 
   <td>1</td> 
   <td>How many potential indicators needed</td> 
  </tr> 
  <tr> 
   <td>crunch_pot_poor_cert</td> 
   <td>-8</td> 
   <td>POTENTIAL crunch cert lt this</td> 
  </tr> 
  <tr> 
   <td>crunch_pot_poor_rate</td> 
   <td>40</td> 
   <td>POTENTIAL crunch rating lt this</td> 
  </tr> 
  <tr> 
   <td>crunch_rating_max</td> 
   <td>10</td> 
   <td>For adj length in rating per ch</td> 
  </tr> 
  <tr> 
   <td>crunch_small_outlines_size</td> 
   <td>0.6</td> 
   <td>Small if lt xht x this</td> 
  </tr> 
  <tr> 
   <td>crunch_terrible_garbage</td> 
   <td>1</td> 
   <td>As it says</td> 
  </tr> 
  <tr> 
   <td>crunch_terrible_rating</td> 
   <td>80</td> 
   <td>crunch rating lt this</td> 
  </tr> 
  <tr> 
   <td>cube_debug_level</td> 
   <td>0</td> 
   <td>Print cube debug info.</td> 
  </tr> 
  <tr> 
   <td>dawg_debug_level</td> 
   <td>0</td> 
   <td>Set to 1 for general debug info, to 2 for more details, to 3 to see all the debug messages</td> 
  </tr> 
  <tr> 
   <td>debug_acceptable_wds</td> 
   <td>0</td> 
   <td>Dump word pass/fail chk</td> 
  </tr> 
  <tr> 
   <td>debug_file</td> 
   <td></td> 
   <td>File to send tprintf output to</td> 
  </tr> 
  <tr> 
   <td>debug_fix_space_level</td> 
   <td>0</td> 
   <td>Contextual fixspace debug</td> 
  </tr> 
  <tr> 
   <td>debug_noise_removal</td> 
   <td>0</td> 
   <td>Debug reassignment of small outlines</td> 
  </tr> 
  <tr> 
   <td>debug_x_ht_level</td> 
   <td>0</td> 
   <td>Reestimate debug</td> 
  </tr> 
  <tr> 
   <td>devanagari_split_debugimage</td> 
   <td>0</td> 
   <td>Whether to create a debug image for split shiro-rekha process.</td> 
  </tr> 
  <tr> 
   <td>devanagari_split_debuglevel</td> 
   <td>0</td> 
   <td>Debug level for split shiro-rekha process.</td> 
  </tr> 
  <tr> 
   <td>disable_character_fragments</td> 
   <td>1</td> 
   <td>Do not include character fragments in the results of the classifier</td> 
  </tr> 
  <tr> 
   <td>doc_dict_certainty_threshold</td> 
   <td>-2.25</td> 
   <td>Worst certainty for words that can be inserted into thedocument dictionary</td> 
  </tr> 
  <tr> 
   <td>doc_dict_pending_threshold</td> 
   <td>0</td> 
   <td>Worst certainty for using pending dictionary</td> 
  </tr> 
  <tr> 
   <td>docqual_excuse_outline_errs</td> 
   <td>0</td> 
   <td>Allow outline errs in unrejection?</td> 
  </tr> 
  <tr> 
   <td>edges_boxarea</td> 
   <td>0.875</td> 
   <td>Min area fraction of grandchild for box</td> 
  </tr> 
  <tr> 
   <td>edges_childarea</td> 
   <td>0.5</td> 
   <td>Min area fraction of child outline</td> 
  </tr> 
  <tr> 
   <td>edges_children_count_limit</td> 
   <td>45</td> 
   <td>Max holes allowed in blob</td> 
  </tr> 
  <tr> 
   <td>edges_children_fix</td> 
   <td>0</td> 
   <td>Remove boxy parents of char-like children</td> 
  </tr> 
  <tr> 
   <td>edges_children_per_grandchild</td> 
   <td>10</td> 
   <td>Importance ratio for chucking outlines</td> 
  </tr> 
  <tr> 
   <td>edges_debug</td> 
   <td>0</td> 
   <td>turn on debugging for this module</td> 
  </tr> 
  <tr> 
   <td>edges_max_children_layers</td> 
   <td>5</td> 
   <td>Max layers of nested children inside a character outline</td> 
  </tr> 
  <tr> 
   <td>edges_max_children_per_outline</td> 
   <td>10</td> 
   <td>Max number of children inside a character outline</td> 
  </tr> 
  <tr> 
   <td>edges_min_nonhole</td> 
   <td>12</td> 
   <td>Min pixels for potential char in box</td> 
  </tr> 
  <tr> 
   <td>edges_patharea_ratio</td> 
   <td>40</td> 
   <td>Max lensq/area for acceptable child outline</td> 
  </tr> 
  <tr> 
   <td>edges_use_new_outline_complexity</td> 
   <td>0</td> 
   <td>Use the new outline complexity module</td> 
  </tr> 
  <tr> 
   <td>editor_dbwin_height</td> 
   <td>24</td> 
   <td>Editor debug window height</td> 
  </tr> 
  <tr> 
   <td>editor_dbwin_name</td> 
   <td>EditorDBWin</td> 
   <td>Editor debug window name</td> 
  </tr> 
  <tr> 
   <td>editor_dbwin_width</td> 
   <td>80</td> 
   <td>Editor debug window width</td> 
  </tr> 
  <tr> 
   <td>editor_dbwin_xpos</td> 
   <td>50</td> 
   <td>Editor debug window X Pos</td> 
  </tr> 
  <tr> 
   <td>editor_dbwin_ypos</td> 
   <td>500</td> 
   <td>Editor debug window Y Pos</td> 
  </tr> 
  <tr> 
   <td>editor_debug_config_file</td> 
   <td></td> 
   <td>Config file to apply to single words</td> 
  </tr> 
  <tr> 
   <td>editor_image_blob_bb_color</td> 
   <td>4</td> 
   <td>Blob bounding box colour</td> 
  </tr> 
  <tr> 
   <td>editor_image_menuheight</td> 
   <td>50</td> 
   <td>Add to image height for menu bar</td> 
  </tr> 
  <tr> 
   <td>editor_image_text_color</td> 
   <td>2</td> 
   <td>Correct text colour</td> 
  </tr> 
  <tr> 
   <td>editor_image_win_name</td> 
   <td>EditorImage</td> 
   <td>Editor image window name</td> 
  </tr> 
  <tr> 
   <td>editor_image_word_bb_color</td> 
   <td>7</td> 
   <td>Word bounding box colour</td> 
  </tr> 
  <tr> 
   <td>editor_image_xpos</td> 
   <td>590</td> 
   <td>Editor image X Pos</td> 
  </tr> 
  <tr> 
   <td>editor_image_ypos</td> 
   <td>10</td> 
   <td>Editor image Y Pos</td> 
  </tr> 
  <tr> 
   <td>editor_word_height</td> 
   <td>240</td> 
   <td>Word window height</td> 
  </tr> 
  <tr> 
   <td>editor_word_name</td> 
   <td>BlnWords</td> 
   <td>BL normalized word window</td> 
  </tr> 
  <tr> 
   <td>editor_word_width</td> 
   <td>655</td> 
   <td>Word window width</td> 
  </tr> 
  <tr> 
   <td>editor_word_xpos</td> 
   <td>60</td> 
   <td>Word window X Pos</td> 
  </tr> 
  <tr> 
   <td>editor_word_ypos</td> 
   <td>510</td> 
   <td>Word window Y Pos</td> 
  </tr> 
  <tr> 
   <td>enable_new_segsearch</td> 
   <td>0</td> 
   <td>Enable new segmentation search path.</td> 
  </tr> 
  <tr> 
   <td>enable_noise_removal</td> 
   <td>1</td> 
   <td>Remove and conditionally reassign small outlines when they confuse layout analysis, determining diacritics vs noise</td> 
  </tr> 
  <tr> 
   <td>equationdetect_save_bi_image</td> 
   <td>0</td> 
   <td>Save input bi image</td> 
  </tr> 
  <tr> 
   <td>equationdetect_save_merged_image</td> 
   <td>0</td> 
   <td>Save the merged image</td> 
  </tr> 
  <tr> 
   <td>equationdetect_save_seed_image</td> 
   <td>0</td> 
   <td>Save the seed image</td> 
  </tr> 
  <tr> 
   <td>equationdetect_save_spt_image</td> 
   <td>0</td> 
   <td>Save special character image</td> 
  </tr> 
  <tr> 
   <td>file_type</td> 
   <td>.tif</td> 
   <td>Filename extension</td> 
  </tr> 
  <tr> 
   <td>fixsp_done_mode</td> 
   <td>1</td> 
   <td>What constitues done for spacing</td> 
  </tr> 
  <tr> 
   <td>fixsp_non_noise_limit</td> 
   <td>1</td> 
   <td>How many non-noise blbs either side?</td> 
  </tr> 
  <tr> 
   <td>fixsp_small_outlines_size</td> 
   <td>0.28</td> 
   <td>Small if lt xht x this</td> 
  </tr> 
  <tr> 
   <td>force_word_assoc</td> 
   <td>0</td> 
   <td>force associator to run regardless of what enable_assoc is.This is used for CJK where component grouping is necessary.</td> 
  </tr> 
  <tr> 
   <td>fragments_debug</td> 
   <td>0</td> 
   <td>Debug character fragments</td> 
  </tr> 
  <tr> 
   <td>fragments_guide_chopper</td> 
   <td>0</td> 
   <td>Use information from fragments to guide chopping process</td> 
  </tr> 
  <tr> 
   <td>fx_debugfile</td> 
   <td>FXDebug</td> 
   <td>Name of debugfile</td> 
  </tr> 
  <tr> 
   <td>gapmap_big_gaps</td> 
   <td>1.75</td> 
   <td>xht multiplier</td> 
  </tr> 
  <tr> 
   <td>gapmap_debug</td> 
   <td>0</td> 
   <td>Say which blocks have tables</td> 
  </tr> 
  <tr> 
   <td>gapmap_no_isolated_quanta</td> 
   <td>0</td> 
   <td>Ensure gaps not less than 2quanta wide</td> 
  </tr> 
  <tr> 
   <td>gapmap_use_ends</td> 
   <td>0</td> 
   <td>Use large space at start and end of rows</td> 
  </tr> 
  <tr> 
   <td>heuristic_max_char_wh_ratio</td> 
   <td>2</td> 
   <td>max char width-to-height ratio allowed in segmentation</td> 
  </tr> 
  <tr> 
   <td>heuristic_segcost_rating_base</td> 
   <td>1.25</td> 
   <td>base factor for adding segmentation cost into word rating.It’s a multiplying factor, the larger the value above 1, the bigger the effect of segmentation cost.</td> 
  </tr> 
  <tr> 
   <td>heuristic_weight_rating</td> 
   <td>1</td> 
   <td>weight associated with char rating in combined cost ofstate</td> 
  </tr> 
  <tr> 
   <td>heuristic_weight_seamcut</td> 
   <td>0</td> 
   <td>weight associated with seam cut in combined cost of state</td> 
  </tr> 
  <tr> 
   <td>heuristic_weight_width</td> 
   <td>1000</td> 
   <td>weight associated with width evidence in combined cost of state</td> 
  </tr> 
  <tr> 
   <td>hocr_font_info</td> 
   <td>0</td> 
   <td>Add font info to hocr output</td> 
  </tr> 
  <tr> 
   <td>hyphen_debug_level</td> 
   <td>0</td> 
   <td>Debug level for hyphenated words.</td> 
  </tr> 
  <tr> 
   <td>il1_adaption_test</td> 
   <td>0</td> 
   <td>Dont adapt to i/I at beginning of word</td> 
  </tr> 
  <tr> 
   <td>include_page_breaks</td> 
   <td>0</td> 
   <td>Include page separator string in output text after each image/page.</td> 
  </tr> 
  <tr> 
   <td>interactive_display_mode</td> 
   <td>0</td> 
   <td>Run interactively?</td> 
  </tr> 
  <tr> 
   <td>language_model_debug_level</td> 
   <td>0</td> 
   <td>Language model debug level</td> 
  </tr> 
  <tr> 
   <td>language_model_fixed_length_choices_depth</td> 
   <td>3</td> 
   <td>Depth of blob choice lists to explore when fixed length dawgs are on</td> 
  </tr> 
  <tr> 
   <td>language_model_min_compound_length</td> 
   <td>3</td> 
   <td>Minimum length of compound words</td> 
  </tr> 
  <tr> 
   <td>language_model_ngram_nonmatch_score</td> 
   <td>-40</td> 
   <td>Average classifier score of a non-matching unichar.</td> 
  </tr> 
  <tr> 
   <td>language_model_ngram_on</td> 
   <td>0</td> 
   <td>Turn on/off the use of character ngram model</td> 
  </tr> 
  <tr> 
   <td>language_model_ngram_order</td> 
   <td>8</td> 
   <td>Maximum order of the character ngram model</td> 
  </tr> 
  <tr> 
   <td>language_model_ngram_rating_factor</td> 
   <td>16</td> 
   <td>Factor to bring log-probs into the same range as ratings when multiplied by outline length</td> 
  </tr> 
  <tr> 
   <td>language_model_ngram_scale_factor</td> 
   <td>0.03</td> 
   <td>Strength of the character ngram model relative to the character classifier</td> 
  </tr> 
  <tr> 
   <td>language_model_ngram_small_prob</td> 
   <td>1e-06</td> 
   <td>To avoid overly small denominators use this as the floor of the probability returned by the ngram model.</td> 
  </tr> 
  <tr> 
   <td>language_model_ngram_space_delimited_language</td> 
   <td>1</td> 
   <td>Words are delimited by space</td> 
  </tr> 
  <tr> 
   <td>language_model_ngram_use_only_first_uft8_step</td> 
   <td>0</td> 
   <td>Use only the first UTF8 step of the given string when computing log probabilities.</td> 
  </tr> 
  <tr> 
   <td>language_model_penalty_case</td> 
   <td>0.1</td> 
   <td>Penalty for inconsistent case</td> 
  </tr> 
  <tr> 
   <td>language_model_penalty_chartype</td> 
   <td>0.3</td> 
   <td>Penalty for inconsistent character type</td> 
  </tr> 
  <tr> 
   <td>language_model_penalty_font</td> 
   <td>0</td> 
   <td>Penalty for inconsistent font</td> 
  </tr> 
  <tr> 
   <td>language_model_penalty_increment</td> 
   <td>0.01</td> 
   <td>Penalty increment</td> 
  </tr> 
  <tr> 
   <td>language_model_penalty_non_dict_word</td> 
   <td>0.15</td> 
   <td>Penalty for non-dictionary words</td> 
  </tr> 
  <tr> 
   <td>language_model_penalty_non_freq_dict_word</td> 
   <td>0.1</td> 
   <td>Penalty for words not in the frequent word dictionary</td> 
  </tr> 
  <tr> 
   <td>language_model_penalty_punc</td> 
   <td>0.2</td> 
   <td>Penalty for inconsistent punctuation</td> 
  </tr> 
  <tr> 
   <td>language_model_penalty_script</td> 
   <td>0.5</td> 
   <td>Penalty for inconsistent script</td> 
  </tr> 
  <tr> 
   <td>language_model_penalty_spacing</td> 
   <td>0.05</td> 
   <td>Penalty for inconsistent spacing</td> 
  </tr> 
  <tr> 
   <td>language_model_use_sigmoidal_certainty</td> 
   <td>0</td> 
   <td>Use sigmoidal score for certainty</td> 
  </tr> 
  <tr> 
   <td>language_model_viterbi_list_max_num_prunable</td> 
   <td>10</td> 
   <td>Maximum number of prunable (those for which PrunablePath() is true) entries in each viterbi list recorded in BLOB_CHOICEs</td> 
  </tr> 
  <tr> 
   <td>language_model_viterbi_list_max_size</td> 
   <td>500</td> 
   <td>Maximum size of viterbi lists recorded in BLOB_CHOICEs</td> 
  </tr> 
  <tr> 
   <td>load_bigram_dawg</td> 
   <td>1</td> 
   <td>Load dawg with special word bigrams.</td> 
  </tr> 
  <tr> 
   <td>load_fixed_length_dawgs</td> 
   <td>1</td> 
   <td>Load fixed length dawgs (e.g. for non-space delimited languages)</td> 
  </tr> 
  <tr> 
   <td>load_freq_dawg</td> 
   <td>1</td> 
   <td>Load frequent word dawg.</td> 
  </tr> 
  <tr> 
   <td>load_number_dawg</td> 
   <td>1</td> 
   <td>Load dawg with number patterns.</td> 
  </tr> 
  <tr> 
   <td>load_punc_dawg</td> 
   <td>1</td> 
   <td>Load dawg with punctuation patterns.</td> 
  </tr> 
  <tr> 
   <td>load_system_dawg</td> 
   <td>1</td> 
   <td>Load system word dawg.</td> 
  </tr> 
  <tr> 
   <td>load_unambig_dawg</td> 
   <td>1</td> 
   <td>Load unambiguous word dawg.</td> 
  </tr> 
  <tr> 
   <td>m_data_sub_dir</td> 
   <td>tessdata/</td> 
   <td>Directory for data files</td> 
  </tr> 
  <tr> 
   <td>matcher_avg_noise_size</td> 
   <td>12</td> 
   <td>Avg. noise blob length</td> 
  </tr> 
  <tr> 
   <td>matcher_bad_match_pad</td> 
   <td>0.15</td> 
   <td>Bad Match Pad (0-1)</td> 
  </tr> 
  <tr> 
   <td>matcher_clustering_max_angle_delta</td> 
   <td>0.015</td> 
   <td>Maximum angle delta for prototype clustering</td> 
  </tr> 
  <tr> 
   <td>matcher_debug_flags</td> 
   <td>0</td> 
   <td>Matcher Debug Flags</td> 
  </tr> 
  <tr> 
   <td>matcher_debug_level</td> 
   <td>0</td> 
   <td>Matcher Debug Level</td> 
  </tr> 
  <tr> 
   <td>matcher_debug_separate_windows</td> 
   <td>0</td> 
   <td>Use two different windows for debugging the matching: One for the protos and one for the features.</td> 
  </tr> 
  <tr> 
   <td>matcher_good_threshold</td> 
   <td>0.125</td> 
   <td>Good Match (0-1)</td> 
  </tr> 
  <tr> 
   <td>matcher_great_threshold</td> 
   <td>0</td> 
   <td>Great Match (0-1)</td> 
  </tr> 
  <tr> 
   <td>matcher_min_examples_for_prototyping</td> 
   <td>3</td> 
   <td>Reliable Config Threshold</td> 
  </tr> 
  <tr> 
   <td>matcher_perfect_threshold</td> 
   <td>0.02</td> 
   <td>Perfect Match (0-1)</td> 
  </tr> 
  <tr> 
   <td>matcher_permanent_classes_min</td> 
   <td>1</td> 
   <td>Min # of permanent classes</td> 
  </tr> 
  <tr> 
   <td>matcher_rating_margin</td> 
   <td>0.1</td> 
   <td>New template margin (0-1)</td> 
  </tr> 
  <tr> 
   <td>matcher_sufficient_examples_for_prototyping</td> 
   <td>5</td> 
   <td>Enable adaption even if the ambiguities have not been seen</td> 
  </tr> 
  <tr> 
   <td>max_permuter_attempts</td> 
   <td>10000</td> 
   <td>Maximum number of different character choices to consider during permutation. This limit is especially useful when user patterns are specified, since overly generic patterns can result in dawg search exploring an overly large number of options.</td> 
  </tr> 
  <tr> 
   <td>max_viterbi_list_size</td> 
   <td>10</td> 
   <td>Maximum size of viterbi list.</td> 
  </tr> 
  <tr> 
   <td>merge_fragments_in_matrix</td> 
   <td>1</td> 
   <td>Merge the fragments in the ratings matrix and delete them after merging</td> 
  </tr> 
  <tr> 
   <td>min_orientation_margin</td> 
   <td>7</td> 
   <td>Min acceptable orientation margin</td> 
  </tr> 
  <tr> 
   <td>min_sane_x_ht_pixels</td> 
   <td>8</td> 
   <td>Reject any x-ht lt or eq than this</td> 
  </tr> 
  <tr> 
   <td>ngram_permuter_activated</td> 
   <td>0</td> 
   <td>Activate character-level n-gram-based permuter</td> 
  </tr> 
  <tr> 
   <td>noise_cert_basechar</td> 
   <td>-8</td> 
   <td>Hingepoint for base char certainty</td> 
  </tr> 
  <tr> 
   <td>noise_cert_disjoint</td> 
   <td>-1</td> 
   <td>Hingepoint for disjoint certainty</td> 
  </tr> 
  <tr> 
   <td>noise_cert_factor</td> 
   <td>0.375</td> 
   <td>Scaling on certainty diff from Hingepoint</td> 
  </tr> 
  <tr> 
   <td>noise_cert_punc</td> 
   <td>-3</td> 
   <td>Threshold for new punc char certainty</td> 
  </tr> 
  <tr> 
   <td>noise_maxperblob</td> 
   <td>8</td> 
   <td>Max diacritics to apply to a blob</td> 
  </tr> 
  <tr> 
   <td>noise_maxperword</td> 
   <td>16</td> 
   <td>Max diacritics to apply to a word</td> 
  </tr> 
  <tr> 
   <td>numeric_punctuation</td> 
   <td>.,</td> 
   <td>Punct. chs expected WITHIN numbers</td> 
  </tr> 
  <tr> 
   <td>ocr_devanagari_split_strategy</td> 
   <td>0</td> 
   <td>Whether to use the top-line splitting process for Devanagari documents while performing ocr.</td> 
  </tr> 
  <tr> 
   <td>ok_repeated_ch_non_alphanum_wds</td> 
   <td>-?*=</td> 
   <td>Allow NN to unrej</td> 
  </tr> 
  <tr> 
   <td>oldbl_corrfix</td> 
   <td>1</td> 
   <td>Improve correlation of heights</td> 
  </tr> 
  <tr> 
   <td>oldbl_dot_error_size</td> 
   <td>1.26</td> 
   <td>Max aspect ratio of a dot</td> 
  </tr> 
  <tr> 
   <td>oldbl_holed_losscount</td> 
   <td>10</td> 
   <td>Max lost before fallback line used</td> 
  </tr> 
  <tr> 
   <td>oldbl_xhfix</td> 
   <td>0</td> 
   <td>Fix bug in modes threshold for xheights</td> 
  </tr> 
  <tr> 
   <td>oldbl_xhfract</td> 
   <td>0.4</td> 
   <td>Fraction of est allowed in calc</td> 
  </tr> 
  <tr> 
   <td>outlines_2</td> 
   <td>ij!?%”:;</td> 
   <td>Non standard number of outlines</td> 
  </tr> 
  <tr> 
   <td>outlines_odd</td> 
   <td>%</td> 
   <td></td> 
  </tr> 
  <tr> 
   <td>output_ambig_words_file</td> 
   <td></td> 
   <td>Output file for ambiguities found in the dictionary</td> 
  </tr> 
  <tr> 
   <td>page_separator</td> 
   <td></td> 
   <td></td> 
  </tr> 
  <tr> 
   <td>pageseg_devanagari_split_strategy</td> 
   <td>0</td> 
   <td>Whether to use the top-line splitting process for Devanagari documents while performing page-segmentation.</td> 
  </tr> 
  <tr> 
   <td>paragraph_debug_level</td> 
   <td>0</td> 
   <td>Print paragraph debug info.</td> 
  </tr> 
  <tr> 
   <td>paragraph_text_based</td> 
   <td>1</td> 
   <td>Run paragraph detection on the post-text-recognition (more accurate)</td> 
  </tr> 
  <tr> 
   <td>permute_chartype_word</td> 
   <td>0</td> 
   <td>Turn on character type (property) consistency permuter</td> 
  </tr> 
  <tr> 
   <td>permute_debug</td> 
   <td>0</td> 
   <td>Debug char permutation process</td> 
  </tr> 
  <tr> 
   <td>permute_fixed_length_dawg</td> 
   <td>0</td> 
   <td>Turn on fixed-length phrasebook search permuter</td> 
  </tr> 
  <tr> 
   <td>permute_only_top</td> 
   <td>0</td> 
   <td>Run only the top choice permuter</td> 
  </tr> 
  <tr> 
   <td>permute_script_word</td> 
   <td>0</td> 
   <td>Turn on word script consistency permuter</td> 
  </tr> 
  <tr> 
   <td>pitsync_fake_depth</td> 
   <td>1</td> 
   <td>Max advance fake generation</td> 
  </tr> 
  <tr> 
   <td>pitsync_joined_edge</td> 
   <td>0.75</td> 
   <td>Dist inside big blob for chopping</td> 
  </tr> 
  <tr> 
   <td>pitsync_linear_version</td> 
   <td>6</td> 
   <td>Use new fast algorithm</td> 
  </tr> 
  <tr> 
   <td>pitsync_offset_freecut_fraction</td> 
   <td>0.25</td> 
   <td>Fraction of cut for free cuts</td> 
  </tr> 
  <tr> 
   <td>poly_allow_detailed_fx</td> 
   <td>0</td> 
   <td>Allow feature extractors to see the original outline</td> 
  </tr> 
  <tr> 
   <td>poly_debug</td> 
   <td>0</td> 
   <td>Debug old poly</td> 
  </tr> 
  <tr> 
   <td>poly_wide_objects_better</td> 
   <td>1</td> 
   <td>More accurate approx on wide things</td> 
  </tr> 
  <tr> 
   <td>preserve_interword_spaces</td> 
   <td>0</td> 
   <td>Preserve multiple interword spaces</td> 
  </tr> 
  <tr> 
   <td>prioritize_division</td> 
   <td>0</td> 
   <td>Prioritize blob division over chopping</td> 
  </tr> 
  <tr> 
   <td>quality_blob_pc</td> 
   <td>0</td> 
   <td>good_quality_doc gte good blobs limit</td> 
  </tr> 
  <tr> 
   <td>quality_char_pc</td> 
   <td>0.95</td> 
   <td>good_quality_doc gte good char limit</td> 
  </tr> 
  <tr> 
   <td>quality_min_initial_alphas_reqd</td> 
   <td>2</td> 
   <td>alphas in a good word</td> 
  </tr> 
  <tr> 
   <td>quality_outline_pc</td> 
   <td>1</td> 
   <td>good_quality_doc lte outline error limit</td> 
  </tr> 
  <tr> 
   <td>quality_rej_pc</td> 
   <td>0.08</td> 
   <td>good_quality_doc lte rejection limit</td> 
  </tr> 
  <tr> 
   <td>quality_rowrej_pc</td> 
   <td>1.1</td> 
   <td>good_quality_doc gte good char limit</td> 
  </tr> 
  <tr> 
   <td>rating_scale</td> 
   <td>1.5</td> 
   <td>Rating scaling factor</td> 
  </tr> 
  <tr> 
   <td>rej_1Il_trust_permuter_type</td> 
   <td>1</td> 
   <td>Dont double check</td> 
  </tr> 
  <tr> 
   <td>rej_1Il_use_dict_word</td> 
   <td>0</td> 
   <td>Use dictword test</td> 
  </tr> 
  <tr> 
   <td>rej_alphas_in_number_perm</td> 
   <td>0</td> 
   <td>Extend permuter check</td> 
  </tr> 
  <tr> 
   <td>rej_trust_doc_dawg</td> 
   <td>0</td> 
   <td>Use DOC dawg in 11l conf. detector</td> 
  </tr> 
  <tr> 
   <td>rej_use_good_perm</td> 
   <td>1</td> 
   <td>Individual rejection control</td> 
  </tr> 
  <tr> 
   <td>rej_use_sensible_wd</td> 
   <td>0</td> 
   <td>Extend permuter check</td> 
  </tr> 
  <tr> 
   <td>rej_use_tess_accepted</td> 
   <td>1</td> 
   <td>Individual rejection control</td> 
  </tr> 
  <tr> 
   <td>rej_use_tess_blanks</td> 
   <td>1</td> 
   <td>Individual rejection control</td> 
  </tr> 
  <tr> 
   <td>rej_whole_of_mostly_reject_word_fract</td> 
   <td>0.85</td> 
   <td>if &gt;this fract</td> 
  </tr> 
  <tr> 
   <td>repair_unchopped_blobs</td> 
   <td>1</td> 
   <td>Fix blobs that aren’t chopped</td> 
  </tr> 
  <tr> 
   <td>save_alt_choices</td> 
   <td>1</td> 
   <td>Save alternative paths found during chopping and segmentation search</td> 
  </tr> 
  <tr> 
   <td>save_doc_words</td> 
   <td>0</td> 
   <td>Save Document Words</td> 
  </tr> 
  <tr> 
   <td>save_raw_choices</td> 
   <td>1</td> 
   <td>Deprecated- backward compatablity only</td> 
  </tr> 
  <tr> 
   <td>segment_adjust_debug</td> 
   <td>0</td> 
   <td>Segmentation adjustment debug</td> 
  </tr> 
  <tr> 
   <td>segment_debug</td> 
   <td>0</td> 
   <td>Debug the whole segmentation process</td> 
  </tr> 
  <tr> 
   <td>segment_nonalphabetic_script</td> 
   <td>0</td> 
   <td>Don’t use any alphabetic-specific tricks.Set to true in the traineddata config file for scripts that are cursive or inherently fixed-pitch</td> 
  </tr> 
  <tr> 
   <td>segment_penalty_dict_case_bad</td> 
   <td>1.3125</td> 
   <td>Default score multiplier for word matches, which may have case issues (lower is better).</td> 
  </tr> 
  <tr> 
   <td>segment_penalty_dict_case_ok</td> 
   <td>1.1</td> 
   <td>Score multiplier for word matches that have good case (lower is better).</td> 
  </tr> 
  <tr> 
   <td>segment_penalty_dict_frequent_word</td> 
   <td>1</td> 
   <td>Score multiplier for word matches which have good case andare frequent in the given language (lower is better).</td> 
  </tr> 
  <tr> 
   <td>segment_penalty_dict_nonword</td> 
   <td>1.25</td> 
   <td>Score multiplier for glyph fragment segmentations which do not match a dictionary word (lower is better).</td> 
  </tr> 
 </tbody> 
</table>

--------------------

[GitHub地址][GitHub]

[1]: /images/20220604/f6a3800478bb422e9f7322968567eb11.png
[2]: /images/20220604/b949968377f24039ae266aca52e3ced4.png
[3]: /images/20220604/851100542dbf46368b68f69bfbabc881.png
[tesseract.js]: http://tesseract.projectnaptha.com/
[SouthEast]: /images/20220604/771b724784034a03bcd2f788f8b79812.png
[ImageLike]: https://github.com/naptha/tesseract.js#imagelike
[GitHub]: https://github.com/yubai8/H5---/tree/master/H5%20%E5%9B%BE%E5%83%8F%E7%AE%80%E5%8D%95%E8%AF%86%E5%88%AB