铁水罐号自动识系统的研究.pdfVIP

摘 要 铁水罐是钢铁冶金重要的物流运输工具，对铁水罐号进行自动识别是企业 生产自动化和管理信息化的基础，开发合理的识别技术将有助于提高我国钢铁 生产自动化和管理信息化的水平。 本文介绍了铁水罐号自动识别系统的研究应用现状，着重讨论了铁水罐定 位、罐号字符区定位、罐号字符切分以及罐号字符识别等关键技术。 对铁水罐定位，本文首先利用改进后的差影法检测出铁水罐的轮廓然后利 用垂直和水平投影定出罐的具体位置。实验结果表明该算法对光照不均匀和复 杂背景并不敏感，能够比较快速、准确、鲁棒地分割出铁水罐部分的图像。 罐号字符区定位是铁水罐号自动识别最关键的一步。基于罐号区域的特 点，本文采用了行扫描法。该算法分为图像预处理、罐号区域搜索、罐号区域 验证三步。利用该算法，本文给出了罐号区域定位部分试验结果。 字符切分是指把罐号字符区域分成单个字符图像，以便进行单个字符识 别。本文在使用基于投影的字符切分法之前还对字符区域进行了连通区域合 并、筛选和倾斜校正等预处理。 字符识别是一个典型的模式识别课题。罐号通常由字母和数字组成，本文 采用了字母和数字分开识别。对于字母，使用简单的统计特征和结构特征区 分；对于数字，使用多重 BP 神经网络进行识别。为进一步提高识别率，本文还 对识别结果做了统计分析。 该系统已经实现，并在国内某大型钢铁企业炼钢厂试用，效果良好。 关键词：计算机视觉，目标定位，字符切分，神经网络，字符识别 I Abstract Molten iron pot (MIP) is an important conveyance in the metallurgy of iron and steel, and for the steel enterprises, the automatic recognition of MIP number (MIPN) is the basis of the automation of production and enterprise management, so the automatic recognition system of molten iron pot number (ARSMIPN) can improve both the automatic level of the production of iron and steel and the enterprise management. In this thesis, current research status and application of ARSMIPN are introduced. And the key techniques including the MIP location, character region location, character segmentation and character recognition are expatiated. For the MIP location, firstly, the algorithm of improved image difference is applied to detect the contour of the MIP, then , a method based on vertical and horizontal projection is designed to calculate the position. The experimental results show this algorithm is efficient in locating the MIP but not sensitive to the unconstrained illumination conditions and irregular background. Character region location is vital in the ARSMIPN. According to the characteri