基于曲线拟合的二次曲面QR码校正

发布时间：2018-05-25 00:35

本文选题：工业QR码 + 自适应伽马校正　；参考：《安徽大学》2017年硕士论文

【摘要】：21世纪,是信息化的社会,信息化社会带给人们更方便快捷的生活方式。图像作为人类感知世界、了解世界的视觉根基,是人类获取信号、表达情感和传递沟通的重要手段,二维码如QR,Data Matrix,Aztec Code,Maxicode等,作为符合以上所有特征的符号,被研发之初主要应用在汽车装配期间汽车零件的跟踪,之后的二十多年间,被扩展应用在各个行业中。除此之外,二维码自身所具备的特点,如:低成本、高可靠性、高保密性、高纠错能力、可编辑外观等,也是其被广泛应用的主要原因。传统流通过程中的QR码多属于半自动的识别,如生活超市中的QR码在第一次扫描失败后,可以更正识别方向继续识别,从而容易检测出符号中蕴含的信息。但工业生产线上的QR码识别,不可轻易改变识别条件,所以识别难度相对较大。本文的主要研究对象是工业二维码。文章介绍了 QR条码图像的基础知识和码制特征以及编译码原理,并对纠错码在解码过程中的步骤有了详细了解。QR码图像识别包括二维码符号锁定,图像清晰处理以及解码等步骤完成,在本文中,研究了二维QR码处理的基本过程,并研究了以下的内容:(1)针对工业QR二维码图像采集环境受限,导致二维码图案质量不佳,包括背景过暗或曝光过强等情况,采用自适应伽马校正方法,增加图像对比度。(2)针对工业QR码图像背景复杂,干扰因素较多,使用双立方插值等算法进行无失真缩放。(3)针对工业QR二维码符号压缩性强,在巨大的零件上占空比小的问题,依据QR二维码自身符号特征,采用图像模板匹配搜索的方式进行二维QR码辅助搜索。在QR码的识别过程中,传统的图像算法对平面图像有很好的处理效果,当这些算法应用在二次曲面失真QR码时识别率就不是理想中的那么好。为此,本文提出—种结合QR码自身符号特征—位置探测图案,基于曲线拟合的曲面QR码校正算法,该算法能够实现对曲面QR码快速精确校正的处理结果。本篇文章整合经典图像预处理的算法以及本文提出的校正算法,使用C++语言编程,在结合了 Google开放源码库ZXing的基础上,设计了针对多种异常环境下采集的图像的软件,经过在本软件以及其他软件上数百张图片的实验,本软件的识别率以及识别时间相比较于其他识别软件有明显的提升。鉴于一般软件无法识别曲面QR码,所以对曲面QR码单独进行实验,验证当图像以不同尺寸处于不同曲率的曲面上时,曲线拟合算法对图像的可识别性。
[Abstract]:In the 21 st century, it is an information society, which brings people a more convenient way of life. As the visual basis of human perception and understanding of the world, image is an important means for human beings to acquire signals, express emotions and transmit communication. QR codes, such as QRX data Matrix Aztec CodeMaxicode, are symbols that conform to all of the above characteristics. It is mainly used in auto parts tracking during automobile assembly at the beginning of R & D, and has been extended to various industries for more than 20 years. In addition, the features of QR codes, such as low cost, high reliability, high confidentiality, high error-correcting ability, editable appearance and so on, are also the main reasons for its wide application. The QR codes in the traditional circulation process are mostly semi-automatic recognition. For example, the QR codes in the life supermarket can correct the recognition direction and continue to recognize after the first scan failure, so that the information contained in the symbol can be easily detected. However, QR code recognition in industrial production line can not easily change the recognition conditions, so it is relatively difficult to identify. The main research object of this paper is industrial two-dimensional code. This paper introduces the basic knowledge of QR barcode image, the character of code system and the principle of encoding and decoding, and also gives a detailed understanding of the steps of error-correcting code in the decoding process, including the two-dimension code symbol lock in the image recognition of .QR code. In this paper, the basic process of two-dimensional QR code processing is studied, and the following contents: 1) aiming at the limitation of the image acquisition environment of industrial QR QR code, the quality of QR code pattern is not good. The background is too dark or the exposure is too strong. The adaptive gamma correction method is used to increase the contrast of the image.) for the industrial QR code image, the background is complex, and the interference factors are many. In order to solve the problem of high compression and low duty cycle of industrial QR QR codes, based on the symbol characteristics of QR QR codes, we use bicubic interpolation and other algorithms to scale without distortion. The image template matching search method is used to carry out two-dimensional QR code aided search. In the process of QR code recognition, the traditional image algorithm has a good processing effect on plane image. When these algorithms are applied to Quadric distortion QR code, the recognition rate is not as good as ideal. Therefore, this paper presents a curved surface QR code correction algorithm based on curve fitting combined with QR code's symbol feature and position detection pattern. The algorithm can realize the fast and accurate correction of surface QR code. This paper integrates the classical image preprocessing algorithm and the correction algorithm proposed in this paper, using C language programming, combined with the Google open source code library ZXing, designed the image acquisition software for a variety of abnormal environments. Through the experiment of hundreds of pictures on this software and other software, the recognition rate and recognition time of this software are obviously improved compared with other recognition software. Because the general software can not recognize the surface QR code, the surface QR code is tested separately to verify the recognition ability of the curve fitting algorithm when the image is on a curved surface with different sizes and different curvature.
【学位授予单位】：安徽大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.4

【参考文献】