复杂背景下的DataMatrix二维码识别算法研究

发布时间：2018-01-11 06:04

本文关键词：复杂背景下的DataMatrix二维码识别算法研究　出处：《深圳大学》2017年硕士论文　论文类型：学位论文

【摘要】：在工业领域,利用二维码对工业产品及零部件进行标识,实现对产品及零部件的生成追踪、装配管理、生命周期维护等已经成为自动化工业的行业标准。其中Data Matrix(DM)二维码因其优秀的数据压缩能力和强大的纠错能力受到工业及物流行业的青睐。与食品、药品和其他消费类产品的包装不同,工业二维码的应用环境通常比较恶劣,二维码的识别通常伴随着噪声、过曝、磨损、污染等问题。尤其是对直接部件表示DPM(Direct Part Marking)的识别,常规的二维码识别方法无法满足其需求。因此针对复杂背景的DM码提取算法具有重要意义和迫切的市场需求。基于上述研究背景和需求,分析了目前工业二维码的技术背景和应用方向,对工业产品及零部件上的DM码开展深入研究,提出了一种基于角度和边缘信息的DM码快速定位和自适应网格划分的DM码采样方法,实现在复杂背景下的DM码快速准确识别。本文的主要研究内容包括以下几个方面:1.根据二维码特征,提出了一种基于角点密度和边缘信息相结合的DM二维码精确定位的方法,包括4部分:基于角点分布的DM码候选区域快速定位和优先级排序;DM码候选区外轮廓提取,减少非感兴趣边缘对定位产生干扰;改进Hough变换粗定位L边,快速进行直线投票;迭代加权最小二乘法直线拟合,精确定位L边。本方法相比传统的Hough变换定位“L”形边方法大幅提高了复杂背景下DM码区域定位的精度、速度和鲁棒性;2.通过分析DM码的特征,提出了基于图像梯度投影累计的自适应采样网格划分算法。通过对DM图像求取梯度,并在其垂直方向上投影累计,将二维图像映射成一维信号,再对该信号进行峰-谷提取,波谷即为采样线位置。最后根据峰谷分布对异常采样线进行校正,从而实现对DM码的自适应采样网格划分。实验结果表明,该方法可有效地对被污染或磨损的DM码进行自适应网格划分,提高恶劣环境下DM码的识别鲁棒性。3.基于嵌入式DSP TMS320DM648搭建DM识别测试系统,以验证本文所提出方法的性能。实验结果表明,所提出的方法对金属材质不同加工工艺的DPM码均具有良好的识别能力,可适应光照不均、低照度、污损等工业复杂环境下的快速、准确读码,满足工业DM识别需求。
[Abstract]:In the field of industry, we use the QR code to mark the industrial products and parts, and realize the generation tracking and assembly management of the products and parts. Life cycle maintenance has become the industry standard for automation industry. QR code is favored by industry and logistics industry because of its excellent data compression ability and powerful error-correcting ability. The packaging of drugs and other consumer products is different. The application environment of industrial QR codes is usually bad. The recognition of QR codes is usually accompanied by noise, overexposure, wear and tear. Problems such as pollution, especially the identification of direct components representing DPM(Direct Part marking. The conventional two-dimension code recognition method can not meet its needs. Therefore, the DM code extraction algorithm for complex background is of great significance and urgent market demand, based on the above research background and requirements. The technical background and application direction of industrial QR codes are analyzed, and the DM codes on industrial products and parts are studied deeply. In this paper, a method of DM code sampling based on angle and edge information is proposed, which is based on fast location of DM code and adaptive trellis division. In this paper, the main research contents include the following aspects: 1. According to the features of two-dimensional code. In this paper, a method of DM two-dimension code precise location based on corner density and edge information is proposed, which includes four parts: fast location and priority ranking for candidate region of DM code based on corner distribution; DM code candidate region contour extraction, reduce the non-interested edge to the location of interference; The improved Hough transform is used to locate the L-edge and to vote in a straight line quickly. The iterative weighted least square method is used to locate the L edge accurately. Compared with the traditional Hough transform method, the accuracy of the region location of DM code in complex background is greatly improved by this method. Speed and robustness; 2. By analyzing the characteristics of DM code, an adaptive sampling mesh generation algorithm based on image gradient projection cumulation is proposed. The gradient of DM image is obtained and the cumulation is projected in its vertical direction. The 2-D image is mapped to one-dimensional signal, and then the peak to valley is extracted, which is the position of the sampling line. Finally, the abnormal sampling line is corrected according to the peak-valley distribution. The experimental results show that the proposed method is effective for adaptive mesh division of DM codes which are contaminated or worn out. Improve the recognition robustness of DM code in bad environment. 3. Build DM recognition test system based on embedded DSP TMS320DM648. In order to verify the performance of the proposed method, the experimental results show that the proposed method has good recognition ability for DPM codes of different processing processes of metal materials, and can adapt to uneven illumination and low illumination. Fast and accurate code reading in complex environment, such as fouling, can meet the requirement of DM recognition.
【学位授予单位】：深圳大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.44

【参考文献】