标签:delete bmp gif htm mic 默认 innertext nes finally
OCR识别推荐两个软件:
Tesseract:一个开源的,由谷歌维护的OCR软件。
Onenote:微软Office附带或者可以自己独立安装。
这次讲Onenote实现的OCR识别。
注:2010版及其以后版本OCR实现方式类似:office将其转换为特定xm格式,然后提取想要的节点就ok了;onenote2007识别比较简单:通过MODI API接口直接之别。
我这里是实现了 office2007和office2010的ocr识别函数。
源程序下载:坚果云连接
1 using Microsoft.Office.Interop.OneNote; 2 using System; 3 using System.Collections.Generic; 4 using System.Drawing; 5 using System.Drawing.Imaging; 6 using System.IO; 7 using System.Linq; 8 using System.Text; 9 using System.Threading.Tasks; 10 using System.Xml; 11 using System.Xml.Linq; 12 13 namespace Extraction.OCR 14 { 15 public class ExtractionOCR 16 { 17 #region 18 private static readonly ExtractionOCR instance = new ExtractionOCR(); 19 public static ExtractionOCR Instance { get { return instance; } } 20 public static string section_path { get; set; } 21 public static int waitTime = 3 * 1000; 22 #endregion 23 /// <summary> 24 /// office2007 MODI组件OCR识别 25 /// </summary> 26 /// <param name="imgPath"></param> 27 /// <returns></returns> 28 public string Ocr_2007(string imgPath) 29 { 30 try 31 { 32 var imgType = imgPath.Substring(imgPath.Length - 3); 33 var data = File.ReadAllBytes(imgPath); 34 string imgInfo = ""; 35 int i = 0; 36 var localimgFile = AppDomain.CurrentDomain.BaseDirectory + @"\" + Guid.NewGuid().ToString() + "." + imgType; 37 while (!imgInfo.Equals("转换成功") && i < 3) 38 { 39 ++i; 40 imgInfo = this.GetBase64(data, imgType, localimgFile); 41 } 42 MODI.Document doc = new MODI.Document(); 43 doc.Create(localimgFile); 44 MODI.Image image; 45 MODI.Layout layout; 46 doc.OCR(MODI.MiLANGUAGES.miLANG_CHINESE_SIMPLIFIED, true, true); 47 StringBuilder sb = new StringBuilder(); 48 image = (MODI.Image)doc.Images[0]; 49 layout = image.Layout; 50 sb.Append(layout.Text); 51 doc = null; 52 var localFilePath = AppDomain.CurrentDomain.BaseDirectory + @"\" + Guid.NewGuid().ToString() + ".txt"; 53 File.WriteAllText(localFilePath, sb.ToString()); 54 Console.WriteLine(sb.ToString()); 55 return localFilePath; 56 } 57 catch (Exception e) 58 { 59 File.AppendAllText(AppDomain.CurrentDomain.BaseDirectory + @"\log.txt", e.ToString()); 60 return ""; 61 } 62 finally 63 { 64 GC.Collect(); 65 } 66 } 67 /// <summary> 68 /// onenote 2010,注意需要先在onenote中创建笔记本,并且将至转换为onenote2007格式 69 /// 推荐使用onenote2016(个人版即可),API与2010类似,(去掉XMLSchema.xs2007参数即可)其他可参考API参数命名。 70 /// 注意1:一定要将dll属性中的“嵌入互操作类型”属性关闭 71 /// </summary> 72 /// <param name="imgPath"></param> 73 /// <returns></returns> 74 public string Ocr_2010(string imgPath) 75 { 76 section_path = @"C:\Users\zhensheng\Desktop\打杂\ocr\ocr.one"; 77 try 78 { 79 if(string.IsNullOrEmpty(section_path)) 80 { 81 Console.WriteLine("请先建立笔记本"); 82 File.AppendAllText(AppDomain.CurrentDomain.BaseDirectory + @"\log.txt", "需要先在onenote中创建笔记本,并且将至转换为onenote2007格式,且将.one文件得路径赋值给section_path"); 83 return ""; 84 } 85 var imgType = imgPath.Substring(imgPath.Length - 3); 86 var data = File.ReadAllBytes(imgPath); 87 string guid = Guid.NewGuid().ToString(); 88 string pageID = ""; 89 string pageXml; 90 XNamespace ns; 91 92 lock (this) 93 { 94 var onenoteApp = new Microsoft.Office.Interop.OneNote.Application(); //onenote提供的API 95 if (onenoteApp == null) 96 { 97 File.AppendAllText(AppDomain.CurrentDomain.BaseDirectory + @"\log.txt", "Microsoft.Office.Interop.OneNote.Application()创建失败"); 98 return ""; 99 } 100 #region 创建页面并返回pageID 101 string sectionID; 102 onenoteApp.OpenHierarchy(section_path, null, out sectionID, CreateFileType.cftSection); 103 pageID = "{" + guid + "}{1}{B0}"; // 格式 {guid}{tab}{??} 104 onenoteApp.CreateNewPage(sectionID, out pageID); 105 #endregion 106 107 #region 获取onenote页面xml结构格式 108 string notebookXml; 109 onenoteApp.GetHierarchy(null, HierarchyScope.hsPages, out notebookXml, XMLSchema.xs2007); 110 var doc = XDocument.Parse(notebookXml); 111 ns = doc.Root.Name.Namespace; 112 113 //var pageNode = doc.Descendants(ns + "Page").FirstOrDefault(); 114 //pageID = pageNode.Attribute("ID").Value; 115 116 #endregion 117 #region 将图片插入页面 118 Tuple<string, int, int> imgInfo = this.GetBase64(data, imgType); 119 var page = new XDocument(new XElement(ns + "Page", 120 new XElement(ns + "Outline", 121 new XElement(ns + "OEChildren", 122 new XElement(ns + "OE", 123 new XElement(ns + "Image", 124 new XAttribute("format", imgType), new XAttribute("originalPageNumber", "0"), 125 new XElement(ns + "Position", 126 new XAttribute("x", "0"), new XAttribute("y", "0"), new XAttribute("z", "0")), 127 new XElement(ns + "Size", 128 new XAttribute("width", imgInfo.Item2), new XAttribute("height", imgInfo.Item3)), 129 new XElement(ns + "Data", imgInfo.Item1))))))); 130 131 page.Root.SetAttributeValue("ID", pageID); 132 onenoteApp.UpdatePageContent(page.ToString(), DateTime.MinValue, XMLSchema.xs2007); 133 #endregion 134 135 #region 通过轮询访问获取OCR识别的结果,轮询超时次数为30次 136 int fileSize = Convert.ToInt32(data.Length / 1024 / 1024); // 文件大小 单位M 137 int count = 0; 138 do 139 { 140 System.Threading.Thread.Sleep(waitTime * (fileSize > 1 ? fileSize : 1)); // 小于1M的都默认1M 141 onenoteApp.GetPageContent(pageID, out pageXml, PageInfo.piBinaryData, XMLSchema.xs2007); 142 ++count; 143 } 144 while (pageXml == "" && count < 6); 145 #endregion 146 147 #region 删除页面 148 onenoteApp.DeleteHierarchy(pageID, DateTime.MinValue); 149 //onenoteApp = null; 150 #endregion 151 } 152 #region 从xml中提取OCR识别后的文档信息,然后输出到string中 153 XmlDocument xmlDoc = new XmlDocument(); 154 xmlDoc.LoadXml(pageXml); 155 XmlNamespaceManager nsmgr = new XmlNamespaceManager(xmlDoc.NameTable); 156 nsmgr.AddNamespace("one", ns.ToString()); 157 XmlNode xmlNode = xmlDoc.SelectSingleNode("//one:Image//one:OCRText", nsmgr); 158 if (xmlNode == null) 159 { 160 File.AppendAllText(AppDomain.CurrentDomain.BaseDirectory + @"\log.txt", "OCR没有识别出值"); 161 return ""; 162 } 163 #endregion 164 var localFilePath = AppDomain.CurrentDomain.BaseDirectory + @"\" + Guid.NewGuid().ToString() + ".txt"; 165 File.WriteAllText(localFilePath, xmlNode.InnerText.ToString()); 166 Console.WriteLine(xmlNode.InnerText.ToString()); 167 168 return localFilePath; 169 } 170 catch (Exception e) 171 { 172 File.AppendAllText(AppDomain.CurrentDomain.BaseDirectory + @"\log.txt", e.ToString()); 173 return ""; 174 } 175 } 176 private string GetBase64(byte[] data, string imgType, string filePath) 177 { 178 using (MemoryStream ms = new MemoryStream()) 179 { 180 MemoryStream ms1 = new MemoryStream(data); 181 Bitmap bp = (Bitmap)Image.FromStream(ms1); 182 switch (imgType.ToLower()) 183 { 184 case "jpg": 185 bp.Save(ms, ImageFormat.Jpeg); 186 break; 187 188 case "jpeg": 189 bp.Save(ms, ImageFormat.Jpeg); 190 break; 191 192 case "gif": 193 bp.Save(ms, ImageFormat.Gif); 194 break; 195 196 case "bmp": 197 bp.Save(ms, ImageFormat.Bmp); 198 break; 199 200 case "tiff": 201 bp.Save(ms, ImageFormat.Tiff); 202 break; 203 204 case "png": 205 bp.Save(ms, ImageFormat.Png); 206 break; 207 208 case "emf": 209 bp.Save(ms, ImageFormat.Emf); 210 break; 211 212 default: 213 return "不支持的图片格式。"; 214 } 215 byte[] buffer = ms.ToArray(); 216 File.WriteAllBytes(filePath, buffer); 217 ms1.Close(); 218 ms.Close(); 219 return "转换成功"; 220 //return new Tuple<string, int, int>(Convert.ToBase64String(buffer), bp.Width, bp.Height); 221 } 222 } 223 private Tuple<string, int, int> GetBase64(byte[] data, string imgType) 224 { 225 using (MemoryStream ms = new MemoryStream()) 226 { 227 MemoryStream ms1 = new MemoryStream(data); 228 Bitmap bp = (Bitmap)Image.FromStream(ms1); 229 switch (imgType.ToLower()) 230 { 231 case "jpg": 232 bp.Save(ms, ImageFormat.Jpeg); 233 break; 234 235 case "jpeg": 236 bp.Save(ms, ImageFormat.Jpeg); 237 break; 238 239 case "gif": 240 bp.Save(ms, ImageFormat.Gif); 241 break; 242 243 case "bmp": 244 bp.Save(ms, ImageFormat.Bmp); 245 break; 246 247 case "tiff": 248 bp.Save(ms, ImageFormat.Tiff); 249 break; 250 251 case "png": 252 bp.Save(ms, ImageFormat.Png); 253 break; 254 255 case "emf": 256 bp.Save(ms, ImageFormat.Emf); 257 break; 258 259 default: 260 return new Tuple<string, int, int>("不支持的图片格式。", 0, 0); 261 } 262 byte[] buffer = ms.ToArray(); 263 ms1.Close(); 264 ms.Close(); 265 return new Tuple<string, int, int>(Convert.ToBase64String(buffer), bp.Width, bp.Height); 266 } 267 } 268 } 269 }
注意:
参考:
标签:delete bmp gif htm mic 默认 innertext nes finally
原文地址:http://www.cnblogs.com/jiangtu/p/6512316.html