Onenote实现OCR识别图片

Posted 2020-09-03 thought

OCR识别推荐两个软件：

1.       Tesseract：一个开源的，由谷歌维护的OCR软件。

2.       Onenote：微软Office附带或者可以自己独立安装。

3.       ONOM：别人封装的onenote api接口

这次讲Onenote实现的OCR识别。github地址：https://github.com/everywan/Extraction.OCR

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（.net4.5，vs2017RC。本机环境：windows10）

 

    

  1 public class ExtractionOCR
  2     {
  3         #region
  4         private static readonly ExtractionOCR instance = new ExtractionOCR();
  5         public static ExtractionOCR Instance { get { return instance; } }
  6         public static string section_path { get; set; }
  7         public static int waitTime = 3 * 1000;
  8         #endregion
  9         /// <summary>
 10         /// office2007 MODI组件OCR识别
 11         /// </summary>
 12         /// <param name="imgPath"></param>
 13         /// <returns></returns>
 14         public string Ocr_2007(string imgPath)
 15         {
 16             try
 17             {
 18                 var imgType = imgPath.Substring(imgPath.Length - 3);
 19                 var data = File.ReadAllBytes(imgPath);
 20                 string imgInfo = "";
 21                 int i = 0;
 22                 var localimgFile = AppDomain.CurrentDomain.BaseDirectory + @"\\" + Guid.NewGuid().ToString() + "." + imgType;
 23                 while (!imgInfo.Equals("转换成功") && i < 3)
 24                 {
 25                     ++i;
 26                     imgInfo = this.GetBase64(data, imgType, localimgFile);
 27                 }
 28                 MODI.Document doc = new MODI.Document();
 29                 doc.Create(localimgFile);
 30                 MODI.Image image;
 31                 MODI.Layout layout;
 32                 doc.OCR(MODI.MiLANGUAGES.miLANG_CHINESE_SIMPLIFIED, true, true);
 33                 StringBuilder sb = new StringBuilder();
 34                 image = (MODI.Image)doc.Images[0];
 35                 layout = image.Layout;
 36                 sb.Append(layout.Text);
 37                 doc = null;
 38                 var localFilePath = AppDomain.CurrentDomain.BaseDirectory + @"\\" + Guid.NewGuid().ToString() + ".txt";
 39                 File.WriteAllText(localFilePath, sb.ToString());
 40                 Console.WriteLine(sb.ToString());
 41                 return localFilePath;
 42             }
 43             catch (Exception e)
 44             {
 45                 File.AppendAllText(AppDomain.CurrentDomain.BaseDirectory + @"\\log.txt", e.ToString());
 46                 return "";
 47             }
 48             finally
 49             {
 50                 GC.Collect();
 51             }
 52         }
 53         /// <summary>
 54         /// 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
 55         /// 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。
 56         /// 注意1：一定要将dll属性中的“嵌入互操作类型”属性关闭
 57         /// </summary>
 58         /// <param name="imgPath"></param>
 59         /// <returns></returns>
 60         public string Ocr_2010(string imgPath)
 61         {
 62             try
 63             {
 64                 #region 确定section_path存在
 65                 section_path = @"C:\\Users\\zhensheng\\Desktop\\打杂\\ocr\\ocr.one";
 66                 if (string.IsNullOrEmpty(section_path))
 67                 {
 68                     Console.WriteLine("请先建立笔记本");
 69                     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");
 70                     return "";
 71                 }
 72                 #endregion
 73                 #region 准备数据
 74                 //后缀
 75                 var imgType = Path.GetExtension(imgPath);
 76                 imgPath = imgPath.Replace(".", "");
 77                 var data = File.ReadAllBytes(imgPath);
 78                 //根据大小确定重试次数
 79                 int fileSize = Convert.ToInt32(data.Length / 1024 / 1024); // 文件大小 单位M
 80                 string guid = Guid.NewGuid().ToString();
 81                 string pageID = "{" + guid + "}{1}{B0}";  // 格式 {guid}{tab}{??}
 82                 string pageXml;
 83                 XNamespace ns;
 84                 var onenoteApp = new Microsoft.Office.Interop.OneNote.Application();  //onenote提供的API
 85                 if (onenoteApp == null)
 86                 {
 87                     File.AppendAllText(AppDomain.CurrentDomain.BaseDirectory + @"\\log.txt", "Microsoft.Office.Interop.OneNote.Application()创建失败");
 88                     return "";
 89                 }
 90                 //重试使用
 91                 XmlNode xmlNode;
 92                 int retry = 0;
 93                 #endregion
 94                 do
 95                 {
 96                     #region 创建页面并返回pageID
 97                     string sectionID;
 98                     onenoteApp.OpenHierarchy(section_path, null, out sectionID, CreateFileType.cftSection);
 99                     onenoteApp.CreateNewPage(sectionID, out pageID);
100                     #endregion
101                     #region 获取onenote页面xml结构格式
102                     string notebookXml;
103                     onenoteApp.GetHierarchy(null, HierarchyScope.hsPages, out notebookXml, XMLSchema.xs2007);
104                     var doc = XDocument.Parse(notebookXml);
105                     ns = doc.Root.Name.Namespace;
106                     #endregion
107                     #region 将图片插入页面
108                     Tuple<string, int, int> imgInfo = this.GetBase64(data, imgType);
109                     var page = new XDocument(new XElement(ns + "Page",
110                                                     new XElement(ns + "Outline",
111                                                     new XElement(ns + "OEChildren",
112                                                         new XElement(ns + "OE",
113                                                         new XElement(ns + "Image",
114                                                             new XAttribute("format", imgType), new XAttribute("originalPageNumber", "0"),
115                                                             new XElement(ns + "Position",
116                                                                 new XAttribute("x", "0"), new XAttribute("y", "0"), new XAttribute("z", "0")),
117                                                             new XElement(ns + "Size",
118                                                                 new XAttribute("width", imgInfo.Item2), new XAttribute("height", imgInfo.Item3)),
119                                                             new XElement(ns + "Data", imgInfo.Item1)))))));
120                     page.Root.SetAttributeValue("ID", pageID);
121                     onenoteApp.UpdatePageContent(page.ToString(), DateTime.MinValue, XMLSchema.xs2007);
122                     #endregion
123                     #region 通过轮询访问获取OCR识别的结果,轮询超时次数为6次
124                     int count = 0;
125                     do
126                     {
127                         System.Threading.Thread.Sleep(waitTime * (fileSize > 1 ? fileSize : 1)); // 小于1M的都默认1M
128                         onenoteApp.GetPageContent(pageID, out pageXml, PageInfo.piBinaryData, XMLSchema.xs2007);
129                     }
130                     while (pageXml == "" && count++ < 6);
131                     #endregion
132                     #region 删除页面
133                     onenoteApp.DeleteHierarchy(pageID, DateTime.MinValue);
134                     //onenoteApp = null;
135                     #endregion
136                     #region 从xml中提取OCR识别后的文档信息，然后输出到string中
137                     XmlDocument xmlDoc = new XmlDocument();
138                     xmlDoc.LoadXml(pageXml);
139                     XmlNamespaceManager nsmgr = new XmlNamespaceManager(xmlDoc.NameTable);
140                     nsmgr.AddNamespace("one", ns.ToString());
141                     xmlNode = xmlDoc.SelectSingleNode("//one:Image//one:OCRText", nsmgr);
142                     #endregion
143                 }
144                 //如果OCR没有识别出信息，则重试三次（个人测试2010失败率为0.2~0.3）
145                 while (xmlNode == null && retry++ < 3);
146                 if (xmlNode == null)
147                 {
148                     File.AppendAllText(AppDomain.CurrentDomain.BaseDirectory + @"\\log.txt", "OCR没有识别出值");
149                     return "";
150                 }
151                 var localFilePath = AppDomain.CurrentDomain.BaseDirectory + @"\\" + Guid.NewGuid().ToString() + ".txt";
152                 File.WriteAllText(localFilePath, xmlNode.InnerText.ToString());
153                 Console.WriteLine(xmlNode.InnerText.ToString());
154                 return localFilePath;
155             }
156             catch (Exception e)
157             {
158                 File.AppendAllText(AppDomain.CurrentDomain.BaseDirectory + @"\\log.txt", e.ToString());
159                 return "";
160             }
161         }
162         private string GetBase64(byte[] data, string imgType, string filePath)
163         {
164             using (MemoryStream ms = new MemoryStream())
165             {
166                 MemoryStream ms1 = new MemoryStream(data);
167                 Bitmap bp = (Bitmap)Image.FromStream(ms1);
168                 switch (imgType.ToLower())
169                 {
170                     case "jpg":
171                         bp.Save(ms, ImageFormat.Jpeg);
172                         break;
173                     case "jpeg":
174                         bp.Save(ms, ImageFormat.Jpeg);
175                         break;
176                     case "gif":
177                         bp.Save(ms, ImageFormat.Gif);
178                         break;
179                     case "bmp":
180                         bp.Save(ms, ImageFormat.Bmp);
181                         break;
182                     case "tiff":
183                         bp.Save(ms, ImageFormat.Tiff);
184                         以上是关于Onenote实现OCR识别图片的主要内容，如果未能解决你的问题，请参考以下文章
 
 图片OCR为啥在线识别不了
 OCR OneNote
 使用Python，几行代码实现OCR图片识别，附测试图片和识别效果
 使用Python，几行代码实现OCR图片识别，附测试图片和识别效果
 项目管理Java OCR实现图片文字识别
 卧槽，javaCV不到十行代码实现图片OCR文字识别