【作者】 陈志强；

【导师】 杨光；

【作者基本信息】 北京邮电大学 ， 机械制造及其自动化， 2013， 硕士

【摘要】 在工业领域中,数字图像处理技术在图像识别中的应用对产品的生产管理、质量的跟踪调查起到了非常重要的作用。传统的图像处理算法为我们解决问题提供了基础,但实际中的图像往往噪声的干扰很大,存在比较严重的降质,用一般方法进行处理难以适应人们的需求,这就要进一步地对相关的处理算法进行深入的研究。本文的研究对象是复杂背景下的工件打码字符图像,对面向字符识别的数字图像预处理及分割算法进行了深入研究,重点是图像增强、图像去噪和图像分割的常用方法。本文结合实际的打码字符识别项目,在相关理论的基础上,设计并实现了一套打码工件字符识别的预处理及分割算法,并对分割后的字符的倾斜矫正和字符在模板中的定位算法进行了研究、设计与实现。实际样本图像的测试结果表明,论文设计的图像处理算法在对目标对象影响较小的情况下,对图像中的大量噪声起到了很好的抑制和去除作用,并且字符能够得到正确的分割。另外,使用本文提出的倾斜矫正算法能够对分割后的字符进行良好的矫正；字符在模板中的定位也较为准确。对论文实现的算法,其整体流程的平均运行时间为700ms左右,其中字符预处理和分割部分为100ms左右。更多还原

【Abstract】 In the industrial field, the application of digital image processing in pattern recognition plays a very important role for production management and quality tracking of products. Traditional image processing algorithms provide basis methods for us to solve problems, but the actual image is often been interfered with much noise and has image degradation. Using normal methods to process image can not adapt to the needs of people, so it is necessary to have a further study on image processing algorithm.The object of this paper is coded character image of work-piece with complex background, the paper studies the digital image pre-processing and segmentation algorithm for character recognition, and focuses on normal methods of image enhancement, image de-noising and image segmentation. Based on the relevant theories, the design and implementation of pre-processing and segmentation algorithm for character recognition are showed in this paper, tilt correction and character positioning algorithms of divided characters are also conducted.The actual test results show that, in the situation of smaller impact on the target object, the image processing algorithm designed in this paper has a good suppression and removal effect on a large amount of noise in the image, and characters can get correct segmentation. In addition, using the tilt correction algorithm mentioned in this paper makes split characters well corrected, and character positioning is also accurate in the template images. The average runtime of the overall algorithm process implemented in this paper is about700ms, of which character pre-processing and segmentation part is about100ms.更多还原