印章图像的自动提取与文字识别(Java,OpenCV)

印章图像的自动提取与文字识别(Java,OpenCV)(论文15000字,程序)
摘要：印章作为正式合同中具有法律效应的有效凭证，在现代经济和社会生活中得到广泛的应用。然而，大多数合同上的印章受文字覆盖、用力不当等多方面因素限制与影响，导致同一印章在每次按压后均表现出不同现象，从而对印章内文字识别产生一定的影响，导致无法判断是否是某公司的有效印章。因此利用图像处理和文字识别等技术对印章自动提取和识别成为一个越来越重要的课题。
以近年来印章提取和识别技术的相关研究成果为基础，对合同印章的高质量提取与印章文字准确识别进行探讨和研究，设计实现了一种针对红色印章的提取方法和一种圆形印章文字识别方法。
本文的主要研究内容如下：
（1）根据RGB颜色空间与HSV颜色空间分布的特点，通过颜色空间转换的组合算法，再根据试验选择合适的阈值，将红色印章提取出来，并尽可能的减少噪声影响。经过多次实验，证明了此种方法可以有效地提取大部分红色印章。
（2）将提取后的二值化印章图片，通过图形切割、坐标转换、畸形矫正，再利用OCR技术识别印章中的文字部分。经过多次实验，其识别正确率可达到76%。
实验结果证明，以上方法较好地解决了复杂文字背景下的红色圆形印章提取难题和印章文字识别问题。 [资料来源：http://www.THINK58.com]
关键词：印章提取；RGB；HSV；图形矫正；OCR文字识别
 
Automatic Extraction and Text Recognition of Seal Images
Abstract：As a valid document with legal effect in the formal contract, the seal is widely used in modern economic and social life. However, most of the seals on the contract are limited and affected by various factors such as text coverage and improper force. As a result, the same seal shows different phenomena after each press, which has a certain impact on the text recognition in the seal, which makes it impossible to judge. Is it a valid seal of a company? Therefore, the use of image processing and text recognition technology to automatically extract and identify seals has become an increasingly important topic.
This study refers to the seal extraction and recognition technology in recent years. It discusses and studies the high-quality extraction of contract seals and the accurate recognition of seal characters. It designs and implements a method for extracting red seals and a method for recognizing circular seal characters.

[资料来源：THINK58.com]

The main research contents of this paper are as follows:
(1) According to the characteristics of RGB color space and HSV color space distribution, through the combination algorithm of color space conversion, select the appropriate threshold according to the experiment, extract the red seal and reduce the noise influence as much as possible. After many experiments, it proved that this method can effectively extract most of the red seals.
(2) The extracted binarized seal picture is image-cut, coordinate-converted, deformed, and OCR technology is used to identify the text portion of the seal. After many experiments, the recognition rate can reach 76%.
The experimental results show that the above method can better solve the problem of red circular seal extraction and seal text recognition under complex text background.
Key words：Seal extraction; RGB; HSV; Graphic Correction; OCR Text Recognition
 

目录
摘要    II
Abstract    III
1引言    1
1.1研究目的及意义    1
1.2研究现状及趋势    1
1.3相关技术介绍    2
1.4本文研究内容    2
2印章提取预处理    3
2.1分割印章区域    3
2.2本章小结    5
3印章提取    5
3.1传统方法    5
3.1.1基于RGB颜色空间的提取方法    5 [资料来源：THINK58.com]
3.1.2基于HSV颜色空间的提取方法    7
3.2基于RGB颜色空间与HSV颜色空间相结合的提取方法    8
3.3相关阈值的调整    9
3.4实验结果与分析    10
3.5图像像素对提取的影响    10
3.6去除噪声影响    12
3.7本章小结    13
4印章文字识别    13
4.1印章文字识别预处理    13
4.1.1极坐标转换    13
4.1.2图形矫正    16
4.2印章文字识别    19
4.3本章小结    20
5总结与展望    20
5.1工作总结    20
5.2改进方向    21
参考文献    22
致谢    23