基于图像识别技术的手写数字识别系统设计(论文11000字)
摘要：
当今社会，全球经济一体化使得阿拉伯数字这一表示财富的字符无时无刻的不出现在身边。当然，数字并不只能用来表示财富，我们生活中每天都要接触运用大量的数字，例如身份证号码、发票数据、数学数字等。不妨设想，如果我们能够运用信息自动化的手段，通过机器去识别处理每天接触的海量的数字信息，社会将节省大量的人力物力,生产活动的效率将大大提高。如果能设计这样的数字识别系统，对社会将会有很高的经济效益。除了具有很高的实际应用价值，手写体数字识别在模式识别领域中同样是一个重要问题，也有着重要的理论价值。
本文提出了一种基于9个特征矢量的特征提取方法，应用了模板匹配知识库的识别算法，给出了一个应用实例：学生成绩单中的手写数字识别。主要工作是对图像进行预处理，主要有图像二值化、数字提取、归一化、最小值滤波、细化等图像处理方法。利用MATLAB函数提取成绩单中的数字,然后选择和数字模板匹配方法的特点的结构特征可以用来识别数字。
关键词:数字识别，特征提取，知识库，模板匹配
Design of The System of Hand-written Numeral Recognition Based on Image Recognition
Abstract:
Today's society, the global economic integration makes the Arabic numerals that the wealth of characters all the time does not appear in the side. Of course, numbers can not only be used to express wealth, we live every day to contact the use of a large number of numbers, such as identity card number, invoice data, mathematical figures. It may be envisaged that if we can use information automation means, through the machine to identify the daily contact with the massive amount of digital information, society will save a lot of manpower and material resources, production efficiency will be greatly improved. If you can design such a digital identification system, the community will have a high economic efficiency. In addition to the high practical application value, handwritten numeral recognition is also an important problem in the field of pattern recognition, and also has important theoretical value.
In the third chapter, we propose a feature extraction and recognition algorithm based on nine feature dimensions, and give an application example: handwritten numeral recognition in student transcripts. The main work is to pretreat the image, mainly image binarization, digital extraction, smoothing filtering, normalization, minimum filtering, refinement and other image processing methods. The use of MATLAB function to extract the figures in the transcripts, and then select the structural features of the features of the digital template matching method can be used to identify the numbers.
Key words:digital recognition, feature extraction, knowledge base, template matching
目 录
一、绪论 1
1.1研究背景及理论意义 1
1.2识别系统性能的评价 1
1.3研究现状 2
1.4本文的研究工作 2
二、数字识别算法 4
2.1基于BP神经网络的识别算法 4
2.2基于结构特征的识别算法 6
2.2.1 Rapid变换系数法 6
2.2.2图像的不变性特征量提取 6
2.2.3方向统计特征量的提取 7
三、手写体数字识别 7
3.1方案设计 7
3.2提取数字 8
3.2.1图像预处理 9
3.2.2数字分离提取 9
3.3识别数字 10
3.3.1细化 10
3.3.2统一规格化 11
3.3.3再细化 11
3.3.4特征提取 11
3.3.5知识库构建 12
3.3.6模板匹配 13
3.4小结 13
四、 全文总结与展望 15
参考文献 16
致谢 18 |
